GPT-5.3 Codex
openai/gpt-5.3-codex- OpenAI
- Family · GPT-5 Codex
- Released 5 Feb 2026
- Text · Image · File → Text
- Reasoning · adjustable
- Tools · JSON · Vision
Spec & list price
Context391K
Max output128K
Input / 1M$1.75
Output / 1M$14.008× input
Scores · standing
Intelligence32.5#59/489
Usage
Tokens / day2.1B▼ 20% 7d
Usage rank#153
Requests / day22.4K
Tool calls / req0.83
Per request · 30 days
Avg input51.1K3.4× median
Avg output5790.8× median
Cache hit91%1.3× median
Reasoning share38%0.7× median
Tool-call failures0.1%0.0× median
Consumption
Peak 40.7B 3 OctLow 1.2B 1 NovRank today #153
News & trends
News · latest from OpenAI
- GPT-6.1 Terra vs. Claude Opus 5.5: Cheaper and faster in three tests at maxthenewstack.io · 2 Nov
- Chinese labs now target OpenAI's GPT-6 Astra for distillationfork.zone · 1 Nov
- GPT-6.1 Terrasimonwillison.net · 31 Oct
- GPT-5.6 Sol can beat Factorio on peaceful moderyanmadden.net · 31 Oct
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 2 providers, latest day
Output speed#143/33474 tok/s1.1× faster than median
Fastest provider90 tok/sbest of 2 providers
First token#2892.91 s2.8× longer than median
First answer token#8036 sthinking included
#9 cheapest of 12 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases32.5
Technical details
- Reasoning
- on · medium
- Tokenizer
- GPT
- Moderation
- moderated
Reasoning effort
- xhigh
- high
- medium · default
- low
- none
Request parameters 11
- tools
- tool_choice
- response_format
- structured_outputs
- reasoning
- reasoning_effort
- include_reasoning
- max_tokens
- max_completion_tokens
- seed
+1 more
- verbosity
Interesting facts
- Weekdays average 6.2B tokens a day against 8.9B on weekends.
- The most used of 3 GPT-5 Codex models by tokens last month.
- Released 270 days ago, the newest of 4 GPT-5 Codex models.
When to use
Best for
- Coding agents that edit large codebases
- Terminal and command-line automation
- Hard debugging with deep reasoning
- Tool-calling developer workflows
- Screenshot and image-aware coding
Think twice for
- General chat and writing
- Low-cost bulk jobs
- New projects (usage is declining)
- Self-hosting (closed weights)
Research papers
- Benchmarks Are Surprisingly Compaction-Friendly22 Oct
- CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents22 Sep
- SkillHEX: Improving Agent Skills via Hypothesis-Driven Autonomous Exploration and Exploitation6 Aug
- IssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests22 Jul
- Specification Grounding Drives Test Effectiveness for LLM Code7 Jul
- SCDBench: A Benchmark for LLM-Based Smart Contract Decompilers27 May
- LivePI: More Realistic Benchmarking of Agents Against Indirect Prompt Injection18 May
- AgentModernize: Preserving Business Logic in Legacy Modernization with Multi-Agent LLMs and Behavioral Specification Graphs17 May
Same family
GPT-5 Codex- 24.9
- 28.5
- 23.7
Alternatives
other vendors · similar capability- 36.0
- 30.1
- 29.7
- 34.0
Using GPT-5.3 Codex at volume?
Sign in to ask for a volume discount. You will continue in the OSMEXO dashboard.