API reference

Every endpoint, parameter, and response.

Base URL
https://artificialanalysis.ai/api/v2
Authentication
API key in x-api-key

Language Models

Evaluations, pricing, performance, and metadata for language models

Get language models (Free tier)

get/api/v2/language/models/free

Free-tier sibling of /api/v2/language/models. Returns the Free-shape language model response: the public subset of fields. Includes headline indices (Intelligence, Coding, Agentic), the six Capability Indexes, median performance, Performance v3 headline means, and input/output pricing. Excludes Pro-only fields such as the full evaluation set, token counts, blended pricing, performance percentiles, time per task and per-workload Performance v3 metrics, context window, parameters, modalities, licensing, and provider detail.

Accepts any valid API key (Free, Pro, or Commercial). Pro and Commercial keys also receive the Free-shape body when calling this URL; the tier field reflects the caller's actual tier.

Parameters
Query parameters
  • pageintegerdefault: 1

    Page number for paginated results (1-indexed). Defaults to 1.

Responses
200
  • tierstringRequired
    freeprocommercial
  • intelligence_index_versionnumberRequired

    Version of the Artificial Analysis Intelligence Index the artificial_analysis_intelligence_index scores in this response belong to, as a number with the v dropped (e.g. 5 for v5.0). Being a number, it carries major.minor only — patch versions are not reflected here. See the Intelligence Index versioning section for what major, minor, and patch changes mean.

  • paginationobjectRequired
    • pageintegerRequired

      Current page number (1-indexed).

    • page_sizeintegerRequired

      Number of items per page.

    • total_pagesintegerRequired

      Total number of pages.

    • has_morebooleanRequired

      Whether there are more pages after the current one.

  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • release_datestring?Required

      ISO date string (e.g. "2024-05-13").

    • model_creatorobjectRequired
    • evaluationsobjectRequired

      Composite and capability Artificial Analysis indices. Pro tier adds per-benchmark scores.

    • artificial_analysis_intelligence_index_costobjectRequired

      Pre-calculated cost to run the Intelligence Index evaluation. Free tier exposes total cost plus cost_per_task.total_cost.

    • pricingobjectRequired

      Input and output token pricing. Pro tier adds blended pricing fields.

    • performanceDeprecatedobjectRequired

      Every field returns null from December 10, 2026. Use performance_v3 (methodology). Median performance metrics across all providers serving this model. Pro tier adds percentiles and quartiles.

    • performance_v3objectRequired

      Performance v3 headline means. Pro tier adds time per task, per-workload metrics and the window date. Null when the model has no measurements in the latest window.

Get all language models

get/api/v2/language/models

Returns language model data including evaluations (headline indices, the six Capability Indexes, and per-benchmark scores), pricing, and performance metrics. Cyber Index scores come with trusted_access: true marks a security model that is not publicly available and has intentionally fewer safeguards. Detail-only fields (evaluation token counts, omniscience/openness breakdowns, providers) are available on /api/v2/language/models/{slug}.

Pro+ only. Free integrations should use /api/v2/language/models/free instead.

Parameters
Query parameters
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
  • pageintegerdefault: 1

    Page number for paginated results (1-indexed). Defaults to 1.

Responses
200
  • tierstringRequired
    procommercial
  • intelligence_index_versionnumberRequired

    Version of the Artificial Analysis Intelligence Index the artificial_analysis_intelligence_index scores in this response belong to, as a number with the v dropped (e.g. 5 for v5.0). Being a number, it carries major.minor only — patch versions are not reflected here. See the Intelligence Index versioning section for what major, minor, and patch changes mean.

  • intelligence_index_evaluationsarray<string>Required

    Keys of the evaluations fields that make up the Intelligence Index at intelligence_index_version. The list changes when the index composition changes, so read it from the response instead of hardcoding it.

  • paginationobjectRequired
    • pageintegerRequired

      Current page number (1-indexed).

    • page_sizeintegerRequired

      Number of items per page.

    • total_pagesintegerRequired

      Total number of pages.

    • has_morebooleanRequired

      Whether there are more pages after the current one.

  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • release_datestring?Required

      ISO date string (e.g. "2024-05-13").

    • model_creatorobjectRequired
    • reasoning_modelboolean

      Pro tier only. Omitted on Free tier.

    • trusted_accessboolean

      True for a model that is not publicly available: a security model with intentionally fewer safeguards, offered only through its creator's trusted-access program. Its Cyber Index scores are not directly comparable with publicly available models. Pro tier only. Omitted on Free tier.

    • evaluationsobjectRequired
    • artificial_analysis_intelligence_index_costobjectRequired

      Pre-calculated cost to run the Intelligence Index evaluation. Free tier includes total_cost and cost_per_task.total_cost; Pro tier adds input_cost, reasoning_cost, answer_cost at both levels.

    • artificial_analysis_intelligence_index_token_countsobject

      Pro tier only. Omitted on Free tier.

    • pricingobject

      Full pricing object including blended fields. Free tier returns input and output token prices only. See FreePricing.

    • performanceDeprecatedobject

      Every field returns null from December 10, 2026. Use performance_v3 (methodology). Full performance object including quartiles and percentiles. Free tier returns medians only. See FreeModelPerformance.

    • performance_v3objectRequired

      Full performance v3 block including time per task, per-workload metrics and the window date. Free tier returns the headline means only. See FreeModelPerformanceV3. Null when the model has no measurements in the latest window.

    • context_window_tokensinteger?

      Pro tier only. Omitted on Free tier. Null when not set for the model.

    • parametersobject

      Pro tier only. Omitted on Free tier.

    • modalitiesobject

      Pro tier only. Omitted on Free tier.

    • licensingobject

      Pro tier only. Omitted on Free tier.

    • huggingface_urlstring?

      Pro tier only. Omitted on Free tier. Canonical Hugging Face or model weights URL when available.

    • openrouter_api_idstring?

      Pro tier only. Omitted on Free tier. OpenRouter API model identifier when available.

    • evaluation_token_countsobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • aa_omniscience_breakdownobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • artificial_analysis_openness_index_breakdownobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • providersarray<object>

      Commercial tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free and Pro tiers and on the list endpoint.

Get a single language model by slug

get/api/v2/language/models/{slug}

Returns data for a single language model identified by its slug. Includes all pro tier list fields plus detail-only fields: evaluation token counts and omniscience/openness breakdowns.

Pro and Commercial only. Free tier requests return a 403 subscription error.

The per-provider breakdown (providers array) is Commercial only and omitted from Pro responses. Contact us to discuss Commercial access.

Parameters
Path parameters1 required
  • slugstringRequired

    The model slug (e.g. "gemini-3-pro", "claude-opus-4-6", "deepseek-v3").

Query parameters
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
Responses
200
  • tierstringRequired
    procommercial
  • intelligence_index_versionnumberRequired

    Version of the Artificial Analysis Intelligence Index the artificial_analysis_intelligence_index scores in this response belong to, as a number with the v dropped (e.g. 5 for v5.0). Being a number, it carries major.minor only — patch versions are not reflected here. See the Intelligence Index versioning section for what major, minor, and patch changes mean.

  • intelligence_index_evaluationsarray<string>Required

    Keys of the evaluations fields that make up the Intelligence Index at intelligence_index_version. The list changes when the index composition changes, so read it from the response instead of hardcoding it.

  • dataobjectRequired
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • release_datestring?Required

      ISO date string (e.g. "2024-05-13").

    • model_creatorobjectRequired
    • reasoning_modelboolean

      Pro tier only. Omitted on Free tier.

    • trusted_accessboolean

      True for a model that is not publicly available: a security model with intentionally fewer safeguards, offered only through its creator's trusted-access program. Its Cyber Index scores are not directly comparable with publicly available models. Pro tier only. Omitted on Free tier.

    • evaluationsobjectRequired
    • artificial_analysis_intelligence_index_costobjectRequired

      Pre-calculated cost to run the Intelligence Index evaluation. Free tier includes total_cost and cost_per_task.total_cost; Pro tier adds input_cost, reasoning_cost, answer_cost at both levels.

    • artificial_analysis_intelligence_index_token_countsobject

      Pro tier only. Omitted on Free tier.

    • pricingobject

      Full pricing object including blended fields. Free tier returns input and output token prices only. See FreePricing.

    • performanceDeprecatedobject

      Every field returns null from December 10, 2026. Use performance_v3 (methodology). Full performance object including quartiles and percentiles. Free tier returns medians only. See FreeModelPerformance.

    • performance_v3objectRequired

      Full performance v3 block including time per task, per-workload metrics and the window date. Free tier returns the headline means only. See FreeModelPerformanceV3. Null when the model has no measurements in the latest window.

    • context_window_tokensinteger?

      Pro tier only. Omitted on Free tier. Null when not set for the model.

    • parametersobject

      Pro tier only. Omitted on Free tier.

    • modalitiesobject

      Pro tier only. Omitted on Free tier.

    • licensingobject

      Pro tier only. Omitted on Free tier.

    • huggingface_urlstring?

      Pro tier only. Omitted on Free tier. Canonical Hugging Face or model weights URL when available.

    • openrouter_api_idstring?

      Pro tier only. Omitted on Free tier. OpenRouter API model identifier when available.

    • evaluation_token_countsobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • aa_omniscience_breakdownobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • artificial_analysis_openness_index_breakdownobject

      Pro tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free tier and list endpoint.

    • providersarray<object>

      Commercial tier only. Detail endpoint only (/api/v2/language/models/{slug}). Omitted on Free and Pro tiers and on the list endpoint.

Get performance over time for a single model

get/api/v2/language/models/{slug}/performance

Returns a time series of performance metrics for the specified model, broken down by API provider. Each provider endpoint that serves this model has its own data points within the requested window.

  • series_v3: Performance v3 for each period: end-to-end output speed, time to first token, time to first answer token and endpoint error rate, as means over the workloads measured (methodology). Data starts on September 15, 2026.
  • series: the older median metrics. Every metric returns null from December 10, 2026. See Deprecations.

Use window to control how far back the data goes and interval to control the granularity (e.g. window=90d&interval=weekly returns ~13 weekly data points per provider).

Commercial only. Free and Pro tier requests return a 403 subscription error. Contact us to discuss Commercial access.

Parameters
Path parameters1 required
  • slugstringRequired

    The unique identifier for the model (e.g. claude-sonnet-4). Use the /models endpoint to discover available slugs.

Query parameters
  • windowstringdefault: "90d"

    Lookback period from today. Maximum 90 days for single-model endpoint.

    7d30d90d
  • intervalstring

    Aggregation granularity. Auto-derived from window if omitted (7d/30d → daily, 90d → weekly).

    dailyweeklymonthly
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
Responses
200
  • tierstringRequired
    commercial
  • dataobjectRequired
    • idstringRequired
    • slugstringRequired
    • namestringRequired
    • providersarray<object>Required

Language Model Providers

Capabilities, pricing, and performance for language model API providers.

Commercial only — these endpoints require a Commercial API key. Contact us to discuss Commercial access.

Get all language model API providers

get/api/v2/language/providers

Returns data for all API providers including the models they offer, pricing, and median performance metrics.

Commercial only. Free and Pro tier requests return a 403 subscription error. Contact us to discuss Commercial access.

Parameters
Query parameters
  • pageintegerdefault: 1

    Page number for paginated results (1-indexed). Defaults to 1.

Responses
200
  • tierstringRequired
    commercial
  • paginationobjectRequired
    • pageintegerRequired

      Current page number (1-indexed).

    • page_sizeintegerRequired

      Number of items per page.

    • total_pagesintegerRequired

      Total number of pages.

    • has_morebooleanRequired

      Whether there are more pages after the current one.

  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • countrystring?

      ISO 3166-1 alpha-2 country code of the provider. Null when unknown.

    • modelsarray<object>Required

Get a single language model provider by slug

get/api/v2/language/providers/{slug}

Returns data for a single API provider identified by its slug. Includes all fields from the providers list plus full performance percentile and quartile data for each model.

Commercial only. Free and Pro tier requests return a 403 subscription error. Contact us to discuss Commercial access.

Parameters
Path parameters1 required
  • slugstringRequired

    The provider slug (e.g. "openai", "fireworks").

Query parameters
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
Responses
200
  • tierstringRequired
    commercial
  • dataobjectRequired
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • countrystring?

      ISO 3166-1 alpha-2 country code of the provider. Null when unknown.

    • modelsarray<object>Required

Get performance over time for a single provider

get/api/v2/language/providers/{slug}/performance

Returns performance time series for each model offered by the specified provider, looking back from today by the specified window.

  • series_v3: Performance v3 for each period: end-to-end output speed, time to first token, time to first answer token and endpoint error rate, as means over the workloads measured (methodology). Data starts on September 15, 2026.
  • series: the older median metrics. Every metric returns null from December 10, 2026. See Deprecations.

This is the inverse of the model performance endpoint — here you pick a provider and see how each of their models has performed over time.

Commercial only. Free and Pro tier requests return a 403 subscription error. Contact us to discuss Commercial access.

Parameters
Path parameters1 required
  • slugstringRequired

    The unique identifier for the provider (e.g. anthropic). Use the /providers endpoint to discover available slugs.

Query parameters
  • windowstringdefault: "90d"

    Lookback period from today.

    7d30d90d
  • intervalstring

    Aggregation granularity. Auto-derived from window if omitted (7d/30d → daily, 90d → weekly).

    dailyweeklymonthly
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
Responses
200
  • tierstringRequired
    commercial
  • dataobjectRequired
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • modelsarray<object>Required

Decision Index

Artificial Analysis Decision Index leaderboard for decision models and language models

Get the Decision Index leaderboard

get/api/v2/language/decision

Returns the Artificial Analysis Decision Index leaderboard: every contestant the public Decision Index page shows, sorted by Decision Index descending. Rows are decision models on their host and, for comparison, language models on an API provider (contestant_kind). Each row carries the index, its headline metrics index and Brier score, each evaluation's Decision Score (or Symmetry Score) and headline metrics, each Decision Score evaluation's Brier score, and the index-weighted cost per task and time per task. For language model rows, model_slug joins to /api/v2/language/models/{slug}. Results are paginated (page).

decision_index and headline_metrics_index are on a 0-100 scale; per-evaluation metrics and Brier scores are on a 0-1 scale. decision_index_version names the index version.

Pro+ only.

Parameters
Query parameters
  • pageintegerdefault: 1

    Page number for paginated results (1-indexed). Defaults to 1.

Responses
200
  • tierstringRequired
    procommercial
  • decision_index_versionstringRequired

    Version of the Artificial Analysis Decision Index the scores in this response belong to, with the v dropped (e.g. 2.1 for v2.1).

  • paginationobjectRequired
    • pageintegerRequired

      Current page number (1-indexed).

    • page_sizeintegerRequired

      Number of items per page.

    • total_pagesintegerRequired

      Total number of pages.

    • has_morebooleanRequired

      Whether there are more pages after the current one.

  • dataarray<object>Required

    Every contestant the public leaderboard shows, decision models and language models, by Decision Index descending.

    • contestant_kindstringRequired

      Whether the row is a decision model on its host, or a language model on an API provider tested for comparison. The site shows language models behind a toggle.

      decision_modellanguage_model
    • slugstringRequired

      Contestant slug, unique within the response. For language_model rows it is the API provider endpoint (host model) slug, not the model slug; see model_slug.

    • model_slugstring?Required

      Slug of the language model behind a language_model row, as used by /api/v2/language/models/{slug}. Null for decision models.

    • namestringRequired

      Contestant name as shown on the site.

    • model_creatorobjectRequired
    • host_namestringRequired

      Name of the host or API provider the model was tested on.

    • decision_indexnumberRequired

      Artificial Analysis Decision Index (0-100): weighted average of the Decision Score on AA-Paperclip (30%), EvalSafe-O*NET-AA (15%), When2Call (10%), AA-Omniscience-Decision (15%) and HLE-MultiChoice (15%), and the Symmetry Score on AA-Symmetry (15%).

    • headline_metrics_indexnumberRequired

      Decision Index headline metrics (0-100): weighted average of each evaluation's headline metrics (the *_all_pass_rate, *_accuracy, *_non_hallucination_rate and *_all_pass_permutation_rate fields) in place of its Decision Score or Symmetry Score. AA-Omniscience-Decision and HLE-MultiChoice each count accuracy at 10% and non-hallucination rate at 5%.

    • brier_scorenumberRequired

      Decision Index Brier score (0-1, lower is better): the five Decision Score evaluations' Brier scores averaged with the index's weights, rescaled to sum to 1.

    • cost_per_tasknumber?Required

      Cost per task in USD at current prices: each evaluation's cost per task averaged with the index's weights. Rounded to 8 decimal places. Null without a cost on every evaluation.

    • time_per_task_secondsnumber?Required

      Seconds per task: each evaluation's median latency averaged with the index's weights. Rounded to 3 decimal places. Null without a latency on every evaluation.

    • aa_paperclip_decision_scorenumberRequired

      AA-Paperclip Decision Score (0-1): (1 + accuracy - Brier score) / 2 of the model's top pick, where 1 is certain and right. A tie counts as the chance a random tie-break is right; an unanswered or errored question as a uniform guess. Each document is one decision, right when every answer is, so accuracy is close to aa_paperclip_all_pass_rate.

    • aa_paperclip_all_pass_ratenumberRequired

      AA-Paperclip all-pass rate (0-1): share of documents with every decision and extraction right.

    • aa_paperclip_brier_scorenumberRequired

      AA-Paperclip Brier score (0-1, lower is better): mean squared gap between each document's probability, the product of its answers' probabilities, and whether every answer is right, as used in the Decision Score.

    • evalsafe_onet_aa_decision_scorenumberRequired

      EvalSafe-O*NET-AA Decision Score (0-1): (1 + accuracy - Brier score) / 2 of the model's top pick, where 1 is certain and right. A tie counts as the chance a random tie-break is right; an unanswered or errored question as a uniform guess. The top pick is the most probable answer. Its accuracy is over all questions, where evalsafe_onet_aa_accuracy averages the three question types.

    • evalsafe_onet_aa_accuracynumberRequired

      EvalSafe-O*NET-AA accuracy (0-1): share of questions where the model's answer is the reference consensus's, averaged over the three question types.

    • evalsafe_onet_aa_brier_scorenumberRequired

      EvalSafe-O*NET-AA Brier score (0-1, lower is better): mean squared gap between the top pick's probability and whether it is right, as used in the Decision Score.

    • when2call_decision_scorenumberRequired

      When2Call Decision Score (0-1): (1 + accuracy - Brier score) / 2 of the model's top pick, where 1 is certain and right. A tie counts as the chance a random tie-break is right; an unanswered or errored question as a uniform guess. The top pick is the most probable reply, so accuracy is close to when2call_accuracy.

    • when2call_accuracynumberRequired

      When2Call accuracy (0-1): share of tasks answered with the correct reply.

    • when2call_brier_scorenumberRequired

      When2Call Brier score (0-1, lower is better): mean squared gap between the top pick's probability and whether it is right, as used in the Decision Score.

    • aa_omniscience_decision_decision_scorenumberRequired

      AA-Omniscience-Decision Decision Score (0-1): (1 + accuracy - Brier score) / 2 of the model's top pick, where 1 is certain and right. A tie counts as the chance a random tie-break is right; an unanswered or errored question as a uniform guess. The top pick is the most probable answer other than "I don't know", so its accuracy can exceed aa_omniscience_decision_accuracy.

    • aa_omniscience_decision_accuracynumberRequired

      AA-Omniscience-Decision accuracy (0-1): share of questions answered correctly.

    • aa_omniscience_decision_non_hallucination_ratenumberRequired

      AA-Omniscience-Decision non-hallucination rate (0-1): share of questions not answered correctly where the model chose a way to say it does not know, or gave no answer.

    • aa_omniscience_decision_brier_scorenumberRequired

      AA-Omniscience-Decision Brier score (0-1, lower is better): mean squared gap between the top pick's probability and whether it is right, as used in the Decision Score.

    • hle_multichoice_decision_scorenumberRequired

      HLE-MultiChoice Decision Score (0-1): (1 + accuracy - Brier score) / 2 of the model's top pick, where 1 is certain and right. A tie counts as the chance a random tie-break is right; an unanswered or errored question as a uniform guess. The top pick is the most probable option other than the declining one, so its accuracy can exceed hle_multichoice_accuracy.

    • hle_multichoice_accuracynumberRequired

      HLE-MultiChoice accuracy (0-1): share of questions answered with the correct letter.

    • hle_multichoice_non_hallucination_ratenumberRequired

      HLE-MultiChoice non-hallucination rate (0-1): share of questions not answered correctly where the model chose the option that declines, or gave no answer.

    • hle_multichoice_brier_scorenumberRequired

      HLE-MultiChoice Brier score (0-1, lower is better): mean squared gap between the top pick's probability and whether it is right, as used in the Decision Score.

    • aa_symmetry_symmetry_scorenumberRequired

      AA-Symmetry Symmetry Score (0-1): per question, 0 if the most probable option changes between forms that mean the same, otherwise 1 less how far that option's probability above chance moves, averaged over questions.

    • aa_symmetry_all_pass_permutation_ratenumberRequired

      AA-Symmetry all-pass permutation rate (0-1): share of questions whose most probable option and probabilities hold across every form.

Media - Image Arena

AA-Image-T2I v2.1 and AA-Image-Editing v2.1 arena rankings for image generation models

Get AA-Image-T2I v2.1 rankings (Free tier)

get/api/v2/media/text-to-image/models/free

Free-tier sibling of /api/v2/media/text-to-image/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Image-T2I v2.1 rankings

get/api/v2/media/text-to-image/models

Returns arena Elo rankings for AA-Image-T2I v2.1, the text-to-image benchmark. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/text-to-image/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_1k_imagesnumber?Required

      Price per 1,000 images.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get AA-Image-Editing v2.1 rankings (Free tier)

get/api/v2/media/image-editing/models/free

Free-tier sibling of /api/v2/media/image-editing/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Image-Editing v2.1 rankings

get/api/v2/media/image-editing/models

Returns arena Elo rankings for AA-Image-Editing v2.1, the image editing benchmark. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/image-editing/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_1k_imagesnumber?Required

      Price per 1,000 images.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get AA-Video-T2V-Silent v2.1 rankings (Free tier)

get/api/v2/media/text-to-video/models/free

Free-tier sibling of /api/v2/media/text-to-video/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Video-T2V-Silent v2.1 rankings

get/api/v2/media/text-to-video/models

Returns arena Elo rankings for AA-Video-T2V-Silent v2.1, the text-to-video benchmark without audio. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/text-to-video/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_minutenumber?Required

      Price per minute of video.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get AA-Video-I2V-Silent v2.1 rankings (Free tier)

get/api/v2/media/image-to-video/models/free

Free-tier sibling of /api/v2/media/image-to-video/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Video-I2V-Silent v2.1 rankings

get/api/v2/media/image-to-video/models

Returns arena Elo rankings for AA-Video-I2V-Silent v2.1, the image-to-video benchmark without audio. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/image-to-video/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_minutenumber?Required

      Price per minute of video.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get AA-Video-T2V v2.1 rankings (Free tier)

get/api/v2/media/text-to-video-audio/models/free

Free-tier sibling of /api/v2/media/text-to-video-audio/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Video-T2V v2.1 rankings

get/api/v2/media/text-to-video-audio/models

Returns arena Elo rankings for AA-Video-T2V v2.1, the text-to-video benchmark for models that generate video with audio output. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/text-to-video-audio/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_minutenumber?Required

      Price per minute of video.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get AA-Video-I2V v2.1 rankings (Free tier)

get/api/v2/media/image-to-video-audio/models/free

Free-tier sibling of /api/v2/media/image-to-video-audio/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Video-I2V v2.1 rankings

get/api/v2/media/image-to-video-audio/models

Returns arena Elo rankings for AA-Video-I2V v2.1, the image-to-video benchmark for models that generate video with audio output. Each model includes identity, Elo and CI, plus rank, samples, pricing, release date, open_weights_url, and optional category breakdowns.

Pro+ only. Free integrations should use /api/v2/media/image-to-video-audio/models/free instead.

Only models that hold a leaderboard rank are returned: current models with at least 500 arena appearances. Models below the threshold and superseded versions are omitted rather than returned with a null rank.

Parameters
Query parameters
  • include_categoriesbooleandefault: false

    Include per-category Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • rankinteger?Required

      Rank among current models by Elo.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_minutenumber?Required

      Price per minute of video.

    • open_weights_urlstring?Required

      URL to open weights; can be used to infer if open weights.

    • categoriesarray<object>

      Per-category Elo breakdowns. Only present when include_categories=true.

Get text-to-speech arena rankings (Free tier)

get/api/v2/media/text-to-speech/models/free

Free-tier sibling of /api/v2/media/text-to-speech/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get text-to-speech arena rankings

get/api/v2/media/text-to-speech/models

Returns arena Elo rankings for text-to-speech models. Each model includes identity, Elo and CI, plus samples, release date, and pricing.

Pro+ only. Free integrations should use /api/v2/media/text-to-speech/models/free instead.

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • samplesintegerRequired

      Total number of arena samples.

    • release_datestring?Required

      Release date.

    • price_per_1m_charactersnumber?Required

      Price per 1 million characters.

Get speech-to-speech model data (Free tier)

get/api/v2/media/speech-to-speech/models/free

Free-tier sibling of /api/v2/media/speech-to-speech/models. Returns the Free-shape response — model identity plus BBA and FDB scores, with Pro-only per-provider pricing and performance data omitted. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • bba_scorenumber?Required

      Big Bench Audio score.

    • fdb_scorenumber?Required

      Full Duplex Bench score.

    • tau_voice_scorenumber?Required

      𝜏-Voice score (the model's best result across providers). Null when the model has no 𝜏-Voice measurements.

Get speech-to-speech model data

get/api/v2/media/speech-to-speech/models

Returns data for speech-to-speech models including quality metrics (BBA and FDB scores). It also includes per-provider pricing and performance data.

Pro+ only. Free integrations should use /api/v2/media/speech-to-speech/models/free instead.

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • model_creatorobjectRequired
    • bba_scorenumber?Required

      Big Bench Audio score.

    • fdb_scorenumber?Required

      Full Duplex Bench score.

    • tau_voice_scorenumber?Required

      𝜏-Voice score (the model's best result across providers). Null when the model has no 𝜏-Voice measurements.

    • providersarray<object>Required

Get speech-to-text models (Free tier)

get/api/v2/media/speech-to-text/models/free

Free-tier sibling of /api/v2/media/speech-to-text/models. Returns the Free-shape response — model identity plus the overall WER index, with Pro-only benchmark breakdowns, open-weights status, and per-provider data omitted. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • aa_wer_indexnumber?Required

      AA Word Error Rate index.

Get speech-to-text models

get/api/v2/media/speech-to-text/models

Returns speech-to-text model data including quality metrics (word error rates). Each model includes identity, the overall WER index, benchmark-specific WER breakdowns, open weights status, and per-provider data.

Pro+ only. Free integrations should use /api/v2/media/speech-to-text/models/free instead.

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • aa_wer_indexnumber?Required

      AA Word Error Rate index.

    • aa_agenttalknumber?Required

      AA AgentTalk WER.

    • voxpopuli_cleaned_aanumber?Required

      VoxPopuli cleaned WER.

    • earnings_22_cleaned_aanumber?Required

      Earnings22 cleaned WER.

    • open_weightsboolean?Required

      Whether the model has open weights. Null when unknown.

    • providersarray<object>Required

      Per-provider data.

Media - Music

AA-Music-Instrumental v1.1 and AA-Music-Vocal v1.1 arena rankings for music generation models

Get AA-Music-Instrumental v1.1 rankings (Free tier)

get/api/v2/media/music/instrumental/models/free

Free-tier sibling of /api/v2/media/music/instrumental/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Music-Instrumental v1.1 rankings

get/api/v2/media/music/instrumental/models

Returns arena Elo rankings for AA-Music-Instrumental v1.1, the instrumental music generation benchmark. Each model includes identity, Elo and CI, plus samples and optional genre breakdowns.

Pro+ only. Free integrations should use /api/v2/media/music/instrumental/models/free instead.

Parameters
Query parameters
  • include_genresbooleandefault: false

    Include per-genre Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • samplesintegerRequired

      Total number of arena samples.

    • genresarray<object>

      Per-genre Elo breakdowns. Only present when include_genres=true.

Get AA-Music-Vocal v1.1 rankings (Free tier)

get/api/v2/media/music/with-vocals/models/free

Free-tier sibling of /api/v2/media/music/with-vocals/models. Returns the Free-shape arena response — model identity plus Elo and CI only. Accepts any valid API key (Free, Pro, or Commercial); Pro and Commercial keys also receive the Free-shape body (Pro-only fields omitted; the tier field reflects the caller's actual tier) when calling this URL.

Responses
200
  • tierstringRequired
    freeprocommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

Get AA-Music-Vocal v1.1 rankings

get/api/v2/media/music/with-vocals/models

Returns arena Elo rankings for AA-Music-Vocal v1.1, the music generation with vocals benchmark. Each model includes identity, Elo and CI, plus samples and optional genre breakdowns.

Pro+ only. Free integrations should use /api/v2/media/music/with-vocals/models/free instead.

Parameters
Query parameters
  • include_genresbooleandefault: false

    Include per-genre Elo breakdowns (Pro tier only).

Responses
200
  • tierstringRequired
    procommercial
  • dataarray<object>Required
    • idstringRequired
    • namestringRequired
    • model_creatorobjectRequired
    • elonumberRequired

      Arena Elo rating.

    • ci_95number?Required

      95% confidence interval. Null if unavailable.

    • samplesintegerRequired

      Total number of arena samples.

    • genresarray<object>

      Per-genre Elo breakdowns. Only present when include_genres=true.

Language Model Measurements

Raw measurement-level data for LLM performance benchmarking.

Commercial only — this endpoint requires a Commercial API key with explicit provider-level grants. Contact us to discuss Commercial access.

Get raw measurements for a provider

get/api/v2/language/providers/{slug}/measurements

Returns measurement records for the specified provider, grouped by model.

  • measurements_v3: one record per Performance v3 run and workload, with median output speed, time to first token, time to first answer token and end-to-end time, plus sample counts.
  • measurements: the older per-request records. Every metric returns null from December 10, 2026. See Deprecations.

Commercial tier only. Your API key must have explicit grants for the requested provider. If you are interested in accessing this endpoint, please contact us to discuss Commercial access.

Parameters
Path parameters1 required
  • slugstringRequired

    The unique identifier for the provider (e.g. anthropic). Your Commercial API key must have grants for this provider.

Query parameters
  • prompt_typeDeprecatedstringdefault: "long"

    Ignored from December 10, 2026.

    The prompt type preset used for performance benchmarking. Controls how many input tokens and parallel queries are used in latency/throughput measurements. Applies to the performance fields only; performance_v3 does not depend on it.

    prompt_typeinput_tokensparallel_queries
    medium1,0001
    long (default)10,0001
    100k100,0001
    vision_single_image1,0001
    medium_parallel1,00010
    mediumlong100kvision_single_imagemedium_parallel
  • start_datestring

    Only return measurements recorded on or after this date (YYYY-MM-DD). Defaults to one month ago if omitted; dates earlier than one month ago are clamped to one month ago.

Responses
200
  • tierstringRequired
    commercial
  • dataobjectRequired
    • idstringRequired
    • namestringRequired
    • slugstringRequired
    • modelsarray<object>Required

CritPt Evaluation

Submit code solutions for grading against the CritPt v2 benchmark.

Unlike the other endpoints in this API, this is not a data retrieval endpoint. It is a code evaluation service. You submit generated code for specific CritPt problems and receive grading results from the CritPt v2 evaluation server.

The switch to the v2 grader does not change the endpoint or request and response schemas.

Access requires approval. Grading is limited to approved labs and researchers — accounts without access receive a 403. Access remains free. To request access, email critpt@artificialanalysis.ai from your institutional address, including the email of the account that holds your API key and details of your research or development use.

Use of this endpoint is subject to our Terms of Use and Data Platform Terms.

Rate limited to 10 requests per day per user. Rate limit headers (X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset, Retry-After) are included in every response.

Submit code for CritPt v2 evaluation

post/api/v2/critpt/evaluate

Submit generated code solutions for grading against CritPt v2 benchmark problems.

This endpoint differs from the rest of the API. It is a code evaluation service, not a data retrieval endpoint. It accepts one or more code submissions and returns grading results from the CritPt v2 evaluation server.

The switch to the v2 grader does not change the endpoint or request and response schemas.

Access requires approval. To protect benchmark integrity and ensure server capacity, grading is limited to approved labs and researchers — accounts without access receive a 403. Access remains free. To request access, email critpt@artificialanalysis.ai from your institutional address, including the email of the account that holds your API key and details of your research or development use.

Use of this endpoint is subject to our Terms of Use and Data Platform Terms.

Rate limited to 10 requests per day per user. Only successful gradings (200) count toward the limit — grader error responses (400/422/503/5xx) and grader timeouts are refunded. Rate limit information is returned via X-RateLimit-* response headers.

Request body
Schema
  • submissionsarray<object>Required

    One or more code submissions to evaluate.

    • problem_idstringRequired

      The CritPt problem identifier.

    • generated_codestringRequired

      The generated code solution to evaluate.

    • modelstringRequired

      The model that generated the code.

    • generation_configobjectRequired

      Configuration used for code generation (passed through to the grader).

    • messagesarray<object>

      Optional conversation messages that produced the code.

  • batch_metadataobjectRequired

    Metadata about the submission batch (passed through to the grader).

Responses
200Evaluation results from the CritPt v2 grading server.