GPT-5.4 (2026-03-05)
Mar 2026
GPT-5.4 (2026-03-05)Compare inon
Benchmarks of GPT-5.4 (2026-03-05)19
BenchmarkByScoreRank
Knowledge2
Reasoning1
Math4
Coding1
Agents2
Legacy · saturated or no longer reported9
No benchmark matches
Rankingat medium effort?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperQwen3.8 Max (0902)xhigh+5%-47%$2 / $6Context: 1.05M → 984KBetterGPT-6 Lunamax+3%-96%$0.068Function calling only with none · Chat Completions+3
OpenDeepSeek V4.1 Flash 1017max+4%-91%$0.26Context: 1.05M → 1M+5
CheaperClaude Sonnet 5.5max+5%-29%$5.46Context: 1.05M → 1M+6
CheaperGemini 4 Argonhigh+5%-29%*$1.62Context: 1.05M → 1M+3ModelScoreCostPer taskWhat changes
Qwen3.8 Max (0902)xhigh+5%-47%$2 / $6Context: 1.05M → 984KClaude Sonnet 5.5max+5%-29%$5.46Context: 1.05M → 1M+6Gemini 4 Argonhigh+5%-29%*$1.62Context: 1.05M → 1M+3GPT-6 Solmax+5%-29%$1.04GPT-6.1 Solmax+5%-29%$0.72Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
GPT-6 Lunamax+3%-96%$0.068Function calling only with none · Chat Completions+3Claude Haiku 5.5xhigh+3%-96%$0.12Context: 1.05M → 1M+6GPT-5.6 Lunamax+3%-92%$0.18DeepSeek V4.1 Flash 1017maxOpen+4%-91%$0.26Context: 1.05M → 1M+5DeepSeek V4 Pro 0813maxOpen+3%-65%$0.67Context: 1.05M → 1M+6ModelScoreCostPer taskWhat changes
DeepSeek V4.1 Flash 1017maxOpen+4%-91%$0.26Context: 1.05M → 1M+5DeepSeek V4 Pro 0813maxOpen+3%-65%$0.67Context: 1.05M → 1M+6
Kimi K3maxOpen+2%+7%$2.00Context: 1.05M → 1M+4Kimi K2.6unstatedOpensame-70%$0.95 / $4Context: 1.05M → 256KKimi K2.7 CodeunstatedOpensame-70%$0.95 / $4Context: 1.05M → 256K+1ModelScoreCostPer taskWhat changes
GPT-6.1 Solmax+5%—$0.72Remove unsupported parameters+5GPT-6 Solmax+5%—$1.04GPT-6 Astramax+5%—$3.26Monitoring can return HTTP 403+2GPT-5.6 Solmax+5%—$1.99GPT-5.6 Terramax+4%—$1.40Prompts with >272K input tokensGPT-6 Lunamax+3%—$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunamax+3%—$0.18GPT-5.2 (2025-12-11)medium-2%—$1.75 / $14Context: 1.05M → 400KAll OpenAI models →History
17 versionsBenchmarks
all categoriesChat & Knowledge
Chess Puzzles37#31
Mystery Game Puzzles28#27
GPQA Diamond · Epoch88.9#38
OTIS Mock AIME 2024 202595.6#32
Coding
SWE-rebench · all · February 1, 2026 to May 15, 202657.6#1
SWE-rebench · Go · February 1, 2026 to May 15, 20260#1
SWE-rebench · Java · February 1, 2026 to May 15, 20260#1
SWE-rebench · Rust · February 1, 2026 to May 15, 20260#1
SWE-rebench · Python · February 1, 2026 to May 15, 202657.6#1
SWE-rebench · TypeScript · February 1, 2026 to May 15, 20260#1
FAQ
How much does GPT-5.4 (2026-03-05) cost?›
$2.50 per 1M input tokens and $15 per 1M output tokens.
Is GPT-5.4 (2026-03-05) free to use?›
There is no free API; the cheapest paid price is $2.50 per 1M input tokens.
How good is GPT-5.4 (2026-03-05), and what is it best at?›
#35 of 84 in Chat & Knowledge, #5 of 24 in Coding, #19 of 31 in Agents & Tool Use.
Is GPT-5.4 (2026-03-05) better than Grok 4.5?›
On OTIS Mock AIME 2024 2025, GPT-5.4 (2026-03-05) scores 97.8 vs 97.8 — about the same.
How can I use GPT-5.4 (2026-03-05)?›
Through the OpenAI API.
What is the context window of GPT-5.4 (2026-03-05)?›
1.05M tokens.
When was GPT-5.4 (2026-03-05) released? Is there a newer version?›
Released 2026-03-05. It is the current version in its line.
Can GPT-5.4 (2026-03-05) read images?›
It accepts text and images as input.
What is GPT-5.4 (2026-03-05)?›
GPT-5.4 (2026-03-05) is a language model from OpenAI, released Mar 2026.
What are the best alternatives to GPT-5.4 (2026-03-05)?›
Qwen3.8 Max (0902) (cheaper), GPT-6 Luna (better), DeepSeek V4.1 Flash 1017 (open weights).