GPT-5 (2025-08-07)
Aug 2025
GPT-5 (2025-08-07)Compare inon
Benchmarks of GPT-5 (2025-08-07)21
BenchmarkByScoreRank
Knowledge2
Reasoning1
Math5Leads on 1
Agents1
Writing2
Legacy · saturated or no longer reported10
No benchmark matches
Rankingat high effort?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGPT-5.6 Lunamax×2.8-87%$0.18BetterGPT-6 Lunamax×2.6-94%$0.068Function calling only with none · Chat Completions+3
OpenKimi K3max+78%+75%$2.00Sampling parameters are fixed+3
CheaperClaude Haiku 5.5max×2.1-94%$0.21Remove sampling parameters+5
CheaperMuse Spark 1.3max×2.1-42%$1.60Only auto tool_choice is supported+5ModelScoreCostPer taskWhat changes
GPT-5.6 Lunamax×2.8-87%$0.18GPT-6 Lunamax×2.6-94%$0.068Function calling only with none · Chat Completions+3Claude Haiku 5.5max×2.1-94%$0.21Remove sampling parameters+5Muse Spark 1.3max×2.1-42%$1.60Only auto tool_choice is supported+5
Qwen3.7 Maxmax+56%-36%$1.48 / $4.43No image inputModelScoreCostPer taskWhat changes
GPT-6 Lunamax×2.6-94%$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunamax×2.8-87%$0.18Claude Haiku 5.5max×2.1-94%$0.21Remove sampling parameters+5
DeepSeek V4 Pro 0813maxOpen+22%-42%$0.67No image input+5GPT-6.1 Solmax×4.6+16%$0.72Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
Kimi K3maxOpen+78%+75%$2.00Sampling parameters are fixed+3
GLM 5.2maxOpen+33%-37%$1.47No image inputGLM 5.3maxOpen+33%-37%$2.01No image input+4DeepSeek V4 Pro 0813maxOpen+22%-42%$0.67No image input+5Kimi K2.6unstatedOpen+17%-50%$0.95 / $4Context: 400K → 256KModelScoreCostPer taskWhat changes
GPT-6.1 Solmax×4.6—$0.72Remove unsupported parameters+5GPT-6 Astrahigh×4.4—$1.73Monitoring can return HTTP 403+2GPT-6 Solmax×4.1—$1.04GPT-5.6 Solmax×3.8—$1.99GPT-5.5 Proxhigh×3.6—$30 / $180GPT-5.5xhigh×3.3—$2.63Prompts with >272K input tokensGPT-5.6 Terramax×3.2—$1.40Prompts with >272K input tokensGPT-5.6 Lunamax×2.8—$0.18All OpenAI models →History
17 versionsBenchmarks
all categoriesChat & Knowledge
SimpleQA Verified v150.1#28
MATH Level 598.1#1
FrontierMath32.4#12
FrontierMath Tier 412.5#10
FrontierMath Tiers 1–3 (v2)55.4#42
Chess Puzzles36#36
Mystery Game Puzzles23#32
GPQA Diamond · Epoch86.2#52
OTIS Mock AIME 2024 202591.4#42
Coding
SWE-bench Pro (public)41.8#8
SWE-rebench · all · 2025-07-01 to 2025-12-0141.9#1
SWE-rebench · Go · 2025-07-01 to 2025-12-010#1
SWE-rebench · Java · 2025-07-01 to 2025-12-010#1
SWE-rebench · Rust · 2025-07-01 to 2025-12-010#1
SWE-rebench · Python · 2025-07-01 to 2025-12-0141.9#1
SWE-rebench · TypeScript · 2025-07-01 to 2025-12-010#1
SWE-bench Verified · Epoch73.6#16
Agents & Tool Use
EBR Bench12.7#19
FAQ
How much does GPT-5 (2025-08-07) cost?›
$1.25 per 1M input tokens and $10 per 1M output tokens.
Is GPT-5 (2025-08-07) free to use?›
There is no free API; the cheapest paid price is $1.25 per 1M input tokens.
How good is GPT-5 (2025-08-07), and what is it best at?›
#28 of 83 in Chat & Knowledge, #8 of 19 in Coding, #19 of 24 in Agents & Tool Use, #33 of 95 in Writing & Language.
Is GPT-5 (2025-08-07) better than Gemini 3.8 Flash?›
On FrontierMath Tier 4 (v2), GPT-5 (2025-08-07) scores 22 vs 22 — about the same.
How can I use GPT-5 (2025-08-07)?›
Through the OpenAI API.
What is the context window of GPT-5 (2025-08-07)?›
400K tokens.
When was GPT-5 (2025-08-07) released? Is there a newer version?›
Released 2025-08-07. It has been superseded by GPT-5.6 Sol.
Can GPT-5 (2025-08-07) read images?›
It accepts text and images as input.
What is GPT-5 (2025-08-07)?›
GPT-5 (2025-08-07) is a language model from OpenAI, released Aug 2025.
What are the best alternatives to GPT-5 (2025-08-07)?›
GPT-5.6 Luna (cheaper), GPT-6 Luna (better), Kimi K3 (open weights).