o1

Dec 2024
o1
Compare in
on
—per task?
$15 / $60per 1M in / outcache $7.50
—output speed?
200Kcontext

Rankingin thinking mode?

Strong at

A
Instructions#49 of 245
B
Tool calling#100 of 242Knowledge#101 of 318
C
Reasoning#222 of 334Science#157 of 312Long context#142 of 273

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperClaude Opus 5.5max×8.8-70%$5.98No forced tool use+5BetterGPT-6 Lunalow×2.9-99%$0.005Function calling only with none · Chat Completions+3OpenMiMo v2.6 Prothinking×7-98%$0.13Pass back reasoning_content+3CheaperClaude Fable 5.1max×8.4-24%$7.63Replace forced tool_choice+5CheaperGemini 4 Argonhigh×8.1-85%*$1.99Synchronous output stops at 65,536+2

History

3 versions
5101520Jan 25Apr 25Release dateHumanity's Last Exam13
5101520Jan 25Apr 25Release dateHumanity's Last Exam13

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam7#224
AA Omniscience Index-11.1#101
CritPt (physics)0.3#149
AA Intelligence Index 4.3.2≈15.2#145
AA Omniscience · Accuracy34.5#62
AA Omniscience · Hallucination Rate69.6#132
AA-LCR (long context) 1.165#142
GPQA Diamond74.7#158
Coding
Terminal-Bench Hard (version unstated)12.9#118
LiveCodeBench · AA archived67.9#55
Agents & Tool Use
τ²-bench · AA62.6#102
Writing & Language
IFBench (instructions)70.3#49

FAQ

How much does o1 cost?›

$15 per 1M input tokens and $60 per 1M output tokens; cached input $7.50 per 1M.

Is o1 free to use?›

There is no free API; the cheapest paid price is $15 per 1M input tokens.

How good is o1, and what is it best at?›

#222 of 334 in Chat & Knowledge, #117 of 239 in Coding, #100 of 242 in Agents & Tool Use, #83 of 195 in Writing & Language, #69 of 88 in Vision & Documents. Strongest at: Instructions.

Is o1 better than GLM 4.5 Air?›

On Humanity's Last Exam, o1 scores 7 vs 7 — about the same.

How can I use o1?›

Through the API of OpenAI.

What is the context window of o1?›

200K tokens, with up to 100K output tokens.

When was o1 released? Is there a newer version?›

Released 2024-12-05. It has been superseded by o3.

Is o1 open source?›

No, o1 is a closed model available through an API. The closest open-weights option here is MiMo v2.6 Pro.

Can o1 read images?›

It accepts text and images, PDF files as input.

What is o1?›

o1 is a closed language model from OpenAI, released Dec 2024.

What are the best alternatives to o1?›

Claude Opus 5.5 (cheaper), GPT-6 Luna (better), MiMo v2.6 Pro (open weights).