whatmodel.sh / GPT-5.3-Codex
GPT-5.3-Codex
↗ share
+ contribute
GPT-5.3-Codex is a closed-weights language model from OpenAI, released February 2026.
$ how it compares
The LLMs ranked either side of GPT-5.3-Codex, out of 828 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 47 |  Grok 4.7 | 75±1 | api $3.00/M | ||
| 48 |  Claude Opus 4.6 | 75±1 | api $10.00/M | ||
| 49 | GPT-5.3-Codex | 74±1 | api $4.81/M | ||
| 50 |  DeepSeek V4 Flash (0731) | 74±1 | api $0.11/M +local | ||
| 51 |  Qwen3.7 Max | 74±1 | api $3.75/M | ||
| more from OpenAI | |||||
| 1 | GPT-6 Astra | 87±1 | api $20.00/M | ||
| 3 | GPT-6.1 Sol | 85±1 | api $4.00/M | ||
| 7 | GPT-6 Sol | 83±1 | api $4.00/M | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #46 of 300 · 99% of its score · 8 tests
Beats Gemini 3.1 Pro (Preview), Muse Spark 1.2, Muse Spark 1.1 here, though they rank higher overall. Trails Claude Sonnet 4.6, Qwen3.8 27B, GLM-5.2 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Verified | 74.8% ✎ | ▁··▁·▁▂▁▄█▄▁ beats 17/29 | 29% |
| BrowseComp lab | 77.3% ✎ | ▄··█▄··▄·▄█▄ beats 0/8 | 20% |
| Toolathlon (Tool Decathlon) lab | 51.9% ✎ | ▃··▃··▃··▃█▅ beats 3/8 | 12% |
| Terminal-Bench 2.0 lab | 77.3% ✎ | ██···██··███ beats 5/6 | 12% |
| Vibe Code Bench (test split, 50 web apps) | 61.8% ✎ | ▃··▃·▃██▃▃▃█ beats 14/14 | 10% |
| OSWorld-Verified lab | 64.7% ✎ | █▄·····▄·▄▄▄ beats 3/6 | 7.1% |
| SWE-bench Pro lab | 56.8% ✎ | ▄·····▄·▄▄▄█ beats 4/6 | 4.7% |
| KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2) | 3.6% ✎ | ▁▁·▁▁▃▁▅▃█▅█ beats 4/27 | 4.5% |
coding▁▁▁▃▆██▇▆▄▃▁ #51 of 458 · 43% of its score · 3 tests
Beats GPT-5.4, DeepSeek V4 Flash Vision (Exp), Gemini 3.1 Pro (Preview) here, though they rank higher overall. Trails Mistral Large 4 (public preview), Qwen3.7 Plus, Claude Sonnet 4.6 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Verified | 74.8% ✎ | ▁··▁·▁▂▁▄█▄▁ beats 17/29 | 29% |
| Vibe Code Bench (test split, 50 web apps) | 61.8% ✎ | ▃··▃·▃██▃▃▃█ beats 14/14 | 10% |
| SWE-bench Pro lab | 56.8% ✎ | ▄·····▄·▄▄▄█ beats 4/6 | 4.7% |
web search▂▁▂▆▄█▃▃▁▁·▁ #34 of 80 * · 20% of its score · 1 test
Beats Gemini 3.1 Pro (Preview) here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BrowseComp lab | 77.3% ✎ | ▄··█▄··▄·▄█▄ beats 0/8 | 20% |
$ where to run it
Through an API
api · $4.81/M · $1.75 in / $14 out per 1M tokens · 400k context · platform.openai.com ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.
$ how the score was worked out · OpenAI vendor-reported benchm… 56%, Epoch AI — internal evaluatio… 29%, Vibe Code Bench (Vals AI, arX… 10% and 1 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 OpenAI vendor-reported benchmark scores for its own models (openai.com announcement posts and Deployment Safety Hub system cards) · lab-reported | 5 | 56% |
 Epoch AI — internal evaluations | 1 | 29% |
 Vibe Code Bench (Vals AI, arXiv:2603.04601v3) | 1 | 10% |
| KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2) | 1 | 4% |
| prior · pull toward the middle | 1% | |
| weighted average | = 74.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json