Skip to content
VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Models741Deals30d269Policy30d65, 14 pendingSources1,014

GPT-5

OpenAIReleased Aug 7, 2025Current

Score basis

GPT-5 is capable in long context, instruction following, and multimodal tasks; and behind the leaders in reasoning, factuality, coding, agentic tasks, and math. Too few results yet to rate safety or multilingual tasks.

Capability profile

Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.

Capable3 capabilities
  1. Long Context−14.5%1 of 3
  2. Instruction Following−18.1%2 of 3
  3. Multimodal−21.6%3 of 6
Limited5 capabilities
  1. Reasoning−27.3%5 of 6
  2. Factuality−29.5%4 of 4
  3. Coding−34.1%8 of 10
  4. Agentic−35.5%3 of 7
  5. Math−41.2%2 of 5
Not rated2 capabilities

Too few results yet to rate Safety or Multilingual.

Price

$1.25input$10.00outputper million tokens

From Artificial Analysis · 2 providers tracked · All prices

Costs more than 78% of 330 priced models · 3:1 input-to-output blend, log scale

Evidence

210results on111benchmarks

  • 53 independently verified
  • 65 aggregator
  • 22 vendor-reported
  • 70 cross-referenced

From 40 sources · latest Oct 8, 2026 · How verification works

API features

Tool callingStructured outputsReasoning

As listed by OpenRouter

Research

856 papers reference GPT-5

GPT-5 benchmark results

210 results on 111 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.

Long Context

CapableFull long context ranking

14.5% behind the leader1 of 3 ranked benchmarks measured

Show 3 more long context resultsHide 3 long context results

Instruction Following

CapableFull instruction following ranking

18.1% behind the leader2 of 3 ranked benchmarks measured

Show 6 more instruction following resultsHide 6 instruction following results

Multimodal

CapableFull multimodal ranking

21.6% behind the leader3 of 6 ranked benchmarks measured

Show 7 more multimodal resultsHide 7 multimodal results

Reasoning

LimitedFull reasoning ranking

27.3% behind the leader5 of 6 ranked benchmarks measured

Show 20 more reasoning resultsHide 20 reasoning results

Factuality

LimitedFull factuality ranking

29.5% behind the leader4 of 4 ranked benchmarks measured

Show 7 more factuality resultsHide 7 factuality results

Coding

LimitedFull coding ranking

34.1% behind the leader8 of 10 ranked benchmarks measured

Show 13 more coding resultsHide 13 coding results

Agentic

LimitedFull agentic ranking

35.5% behind the leader3 of 7 ranked benchmarks measured

Show 4 more agentic resultsHide 4 agentic results

Math

LimitedFull math ranking

41.2% behind the leader2 of 5 ranked benchmarks measured

Show 3 more math resultsHide 3 math results

More results

Benchmarks outside the capability baskets. They are not ranked against a leader.

Show 113 more resultsHide 113 results

GPT-5: common questions

Who makes GPT-5?

GPT-5 is made by OpenAI.

When was GPT-5 released?

GPT-5 was released on Aug 7, 2025, according to Artificial Analysis.

What is GPT-5 good at?

GPT-5 is capable in long context, instruction following, and multimodal tasks; and behind the leaders in reasoning, factuality, coding, agentic tasks, and math. Too few results yet to rate safety or multilingual tasks.

How much does GPT-5 cost?

GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens, according to Artificial Analysis. We track its price at 2 providers. At a mix of three input tokens to one output token, it costs more than 78% of the 330 priced models we track.

How many benchmarks has GPT-5 been tested on?

We track 210 results for GPT-5 on 111 benchmarks from 40 sources, 53 of them independently verified. The latest was recorded on Oct 8, 2026.

Which API features does GPT-5 support?

OpenRouter lists tool calling, structured outputs, and reasoning for GPT-5.

About this record

Where GPT-5's numbers come from, and every name it appears under.

Tracked since
May 1, 2026
Newest source mention
Aug 24, 2026

Where the results come from

Verification: 210 scores · 53 independently verified · 65 aggregator-attributed · 70 vendor cross-reference · 22 vendor-reported. How these tiers are assigned

From 40 sources on 18 sites. Hugging Face supplies 68 of them; the 53 independently verified results come from 14 sites. Bars are coloured by trust tier.

  • huggingface.co68
  • artificialanalysis.ai67
  • api.llm-stats.com18
  • raw.githubusercontent.com10
  • arcprize.org8
  • epoch.ai6
  • matharena.ai6
  • storage.googleapis.com6
  • x.ai5
  • deploymentsafety.openai.com4
  • aider.chat2
  • datasets-server.huggingface.co2
  • labs.scale.com2
  • swebench.com2
  • 99franklin.github.io1
  • lmarena.ai1
  • simple-bench.com1
  • www-cdn.anthropic.com1

Also known as

How our sources name GPT-5 at each reasoning setting.

SettingShort formAPI id
minimal—gpt-5 (minimal) gpt-5-minimal gpt-5-minimal-2025-08-07
low—gpt-5 (low) gpt-5-low
mediumGPT 5 (2025-08-07) (medium)gpt-5 (2025-08-07) (medium reasoning) gpt-5 (medium) gpt-5-medium
highgpt-5 highgpt-5 (high) gpt-5 (high) (Non-Reasoning) gpt-5 (high) (reasoning) gpt-5-2025-08-07 (High) gpt-5-high-2025-08-07
Also listed asgpt-5 (2025-08-07)gpt-5 (inspect)gpt-5 (thinking)gpt-5-thinkinggpt-5:batchgpt5gpt5-2025-08-07gpt-5-2025-08-07gpt-5-2025-08-07-thinking