GPT-5.4
openai/gpt-5.4- OpenAI
- Family · GPT-5
- Released 5 Mar 2026
- Text · Image · File → Text
- Reasoning · adjustable
- Tools · JSON · Vision
Spec & list price
Context1M
Max output128K
Input / 1M$2.50
Output / 1M$15.006× input
Scores · standing
Intelligence39.0#39/489
Coding71.1#25/196
Usage
Tokens / day8.2B▲ 12% 7d
Usage rank#95
Requests / day418.6K
Tool calls / req0.10
Per request · 30 days
Avg input7.4K0.5× median
Avg output2420.4× median
Cache hit68%about median
Reasoning share30%0.5× median
Tool-call failures1.3%0.7× median
Consumption
Peak 13.1B 29 SepLow 3.6B 17 OctRank today #95
News & trends
9 mentions this weekthis model · videos, posts, articles · 30 days
Trending · 14 days · GPT-5 family
- GPT-5.6 Sol can beat Factorio on peaceful mode175 points
- The Current Models: GPT-5.5 & GPT-5.6 (Sol, Terra, Luna) | ChatGPT 2026 (Ep. 4)7.9K views
- @ArtificialAnlys on X43.1K views
- OpenAI researcher defects to Anthropic after shipping GPT-51 upvotes
- [OC] I checked 163M Reddit posts and comments. Em dashes took off after GPT-4o, then fell 17% the month GPT-5 came out403 upvotes
- Codex 逆袭：从失败到 ChatGPT 底座 | OpenAI | ChatGPT | GPT-5 | Claude Code | o1 | 思维链 | 创新者的窘境 | Sam Altman28.1K views
News · latest from OpenAI
- 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
- Post-training 4B model with RL to fly Tesla Roadster 21% better than GPT-6 Astraskypilot.ai · 30 Oct
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 7 providers, latest day
Output speed#120/33482 tok/s1.2× faster than median
Fastest provider103 tok/sbest of 7 providers
First token#2241.54 s1.5× longer than median
#15 cheapest of 18 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases38.4
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
When to use
Best for
- Coding agents with terminal workflows
- Very long document and codebase analysis
- Hard science and reasoning questions
- Tool-calling business workflows
- Offline bulk jobs via the batch variant
Think twice for
- High-volume simple chat (cheaper options exist)
- Self-hosting (closed weights)
- Image generation
Research papers
- Benchmarking color judgment in multimodal LLMs with CIE-computable ground truth29 Oct
- Where You Inject Matters: The Trust Region Effect in LLM Agents28 Oct
- HIDE: Hierarchical Isolation via Documentation and Evidence for Rustc Bugs27 Oct
- SEMA-Auto: Sighted and Efficient Multi-Turn Attack with Autonomy26 Oct
- From Document Layout to Causal Topology: An End-to-End Architecture for Temporal Causal Reasoning25 Oct
- TIDE: Tree-Induced Fine-Grained Diagnostic Evaluation from Model Behaviors21 Oct
- ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination21 Oct
- Decision language models as affordable and rapid alternatives to LLM judges of clinical generative AI20 Oct
Social
Most watched and discussed · 90 days
- The Early Models: From GPT-5 to GPT-5.4 | ChatGPT 2026 (Ep. 3)Snob Bots · 6.9K views · 23 Oct
- GPT-5.4 vs Grok 4.20 | Men O'z AI Agentimni Yasab Oldim!youtrix · 105 views · 27 Oct
- GitHub Copilot removes GPT-5.5, GPT-5.4 and four more modelsJys on tech · 40 views · 29 Oct
- GPT-5.4 Is Here — OpenAI's First Model That Uses Your Computer | NOVA KreationsNOVA Kreations · 26 views · 21 Oct
- AI SHOCKWAVES: GPT-5.4, Microsoft Cuts, Data Crisis, AI Hackers!ShaqTech · 13 views · 27 Oct
- A Red Heart Image Made Claude and GPT-5.4 Write Recursion | #ShortsThe Business Engineer · 9 views · 28 Oct
- GPT-5.4 の正体｜パソコン操作で「人間超え」の中身と、働くAIの手になった意味AIの手の内 · 9 views · 20 Oct
- My Laptop Model Tied GPT-5.4-mini and Was 9× Faster. Is It Cheaper? Self-Hosting vs API, MeasuredSlav Kurochkin · 6 views · 27 Oct
Same family
GPT-5- 23.0
- 38.4
- 30.4
- 24.7
Alternatives
other vendors · similar capability- S43.7
- 36.0
- 38.2
- 41.8
Using GPT-5.4 at volume?
Sign in to ask for a volume discount. You will continue in the OSMEXO dashboard.