OpenAI · language model
o3: pricing, context and benchmarks
o3 is a language model from OpenAI released on April 16, 2025. It costs $2 per million input tokens and $8 per million output tokens, with a 200K-token context window. Epoch AI ranks it #73 of 125 rated models.
What you need to know about o3
o3 was released by OpenAI (United States) on April 16, 2025. It accepts images, text, and files as input and produces text. Its context window reaches 200,000 tokens, larger than 36 % of the models in the registry, and it can write up to 100,000 tokens in one go.
Released more than 1 year ago, it belongs to a generation most vendors have since superseded.
o3 pricing
On OpenRouter, o3 costs $2 per million input tokens and $8 per million output tokens. Counting three tokens read for every token written, its average price is $3.5 per million, more expensive than 81 % of paid models (4.6 times the median of $0.76).
A concrete example: a task that sends 2,000 tokens (about 1,500 words) and receives a 500-token answer costs $0.008. At 1,000 tasks a day, budget $240 a month. The price calculator redoes this with your own volumes.
Tokens already sent and cached are billed at $0.5 per million. With batch processing (delayed answers), the price drops to $1 input and $4 output.
Performance and benchmarks
The Epoch Capabilities Index (ECI), which aggregates published results across many benchmarks, gives o3 a score of 146.7 (range 144.8 to 148.5). That places it #73 of 125 rated models.
Scores published by Epoch AI on the benchmarks the site tracks:
| Benchmark | What it measures | Score | Date |
|---|---|---|---|
| GPQA Diamond | PhD-level science questions | 81.8 % | Apr 16, 2025 |
| SWE-bench Verified | fixing real GitHub issues | 62.3 % | Apr 16, 2025 |
| ARC-AGI-1 | visual abstract-reasoning puzzles | 60.8 % | Apr 16, 2025 |
| ARC-AGI-2 | harder version of the ARC puzzles | 6.5 % | Apr 16, 2025 |
| FrontierMath | research-level maths problems (tiers 1-3) | 33.3 % | Apr 16, 2025 |
| OTIS Mock AIME | maths competition problems | 84.4 % | Apr 16, 2025 |
| SimpleQA Verified | accuracy on simple facts | 49.4 % | Apr 16, 2025 |
| Aider Polyglot | coding exercises in several languages | 81.3 % | Apr 16, 2025 |
| OSWorld | operating a desktop computer | 23 % | Apr 16, 2025 |
For at least equal performance, Claude Haiku 5.5, Gemini 3.7 Flash, and Gemini 3.8 Flash cost at least 30% less.
Features, privacy and local use
- It supports tool calling, so it can drive an AI agent, call APIs or MCP servers.
- It can answer in an enforced JSON format (structured outputs).
- It is a reasoning model: it "thinks" before answering, which helps on hard tasks but uses more output tokens.
- Its knowledge stops at June 2024.
Its weights are not published: it can only be used through the OpenAI API or a broker such as OpenRouter, which means sending your data off your machines.
Other OpenAI models
| Model | Released | Avg. price | ECI |
|---|---|---|---|
| GPT-6.1 Sol | Sep 29, 2026 | $4 | 168.4 |
| GPT-6.1 Sol Pro | Sep 29, 2026 | $4 | — |
| GPT-6 Sol | Sep 22, 2026 | $4 | 163.7 |
| GPT-6 Luna | Sep 22, 2026 | $0.2 | 156.8 |
| GPT-6 Luna Pro | Sep 22, 2026 | $0.2 | — |
| GPT-6 Sol Pro | Sep 22, 2026 | $4 | — |
| GPT-6 Astra | Sep 4, 2026 | $20 | 168.8 |
| GPT-6 Astra Pro | Sep 4, 2026 | $20 | — |
Verdict
o3 scores 39/100 on our grid, which only uses published figures.
Strengths
- Agent-ready: tool calling and structured JSON output.
- Discounted batch pricing.
Limitations
- Expensive: costlier than 81 % of paid models.
- Older generation, often replaced by a newer model from the same vendor.
Avoid if…
- you process large volumes on a tight budget;
- your data must not leave your servers;
- your tasks are mostly simple: a non-reasoning model will cost less.
Alternatives
- GLM 5.1$2.15
- GLM 5$1.55
- Qwen3.5 397B A17B$1.28
- Claude Opus 4.5$10
- Claude Sonnet 4.5$6
- Claude Haiku 4.5$2
Comparisons and tools
Learn more
Frequently asked questions
How much does o3 cost?
$2 per million input tokens and $8 per million output tokens on OpenRouter (checked October 25, 2026). A task with 2,000 tokens read and 500 written costs $0.008.
How big is o3's context window?
200,000 tokens, roughly 150,000 words, with a maximum of 100,000 output tokens.
Can o3 run locally?
No, OpenAI does not publish its weights: it is API-only.
Is o3 suitable for an AI agent?
Technically yes: it supports tool calling and structured outputs. Then check its cost for your volume.
How good is o3?
Epoch AI ranks it #73 of 125 rated models (ECI 146.7). On GPQA Diamond, it scores 81.8%.
What are the alternatives to o3?
The closest models on price and performance are GLM 5.1, GLM 5, and Qwen3.5 397B A17B.
Related articles
- Mistral AI: models, pricing and agentsOctober 25, 2026
- LLM benchmarks: how to read model rankingsOctober 24, 2026
- Running a local LLM: hardware, software, and limitationsOctober 24, 2026
- Open-source or closed models for your AI agents?October 5, 2026
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score