Model Detail
GPT-5.4
Dominates terminal-style external benchmarks, but still has a gap in our structured internal operator benchmark data.
Terminal/coding specialist
- Rank
- #8 of 13
- Source
- External benchmark canon
- Run date
01 · Benchmark score 88/100
02 · Source External benchmark canon
03 · Role Terminal/coding specialist
Per-suite breakdown
Scorecard
Operator board #9
88/100 operator score
Messaging Provider path unsupported
Messaging Tool Planning v2
External canon 75.1 Terminal-Bench · 57.7 SWE-bench Pro · 1463 Arena
Public benchmark references
Cost $8.75/M blended
Looks strongest for coding/execution, still under-benchmarked internally.
Strengths
- #1 on Terminal-Bench 2.0
- Very strong coding/execution profile
- Competitive frontier model
Weaknesses
- Provider path blocked prior internal benchmarking
- Less direct operator-suite evidence in our canon
Dominates terminal-style external benchmarks, but still has a gap in our structured internal operator benchmark data.
Comparison
OperatorIndex Where GPT-5.4 lands
- 01 Gemini 3.1 ProMessaging benchmark + external canon 100
- 01 Claude Sonnet 4.6Messaging benchmark + external canon 100
- 01 Gemini FlashMessaging benchmark canon 100
- 04 Claude Opus 4.6Operator Suite v2 95.3
- 05 GLM-5-TurboOperator Suite v2 95
- 06 Gemma 4 31BEnterprise Ollama 92
- 07 MiniMax M2.7Operator Suite v2 90.8
- 08 GPT-5.4External benchmark canon 88
- 09 Gemma 4 26BEnterprise Ollama 80
- 10 PrismML Bonsai 1.7BPrismML local benchmark 56
- 11 Qwen 3.6 27B NVFP4Enterprise oMLX 40
- 11 Qwen 3.6 27B MXFP4Enterprise oMLX 40
- 13 Qwen 3.6 27B 4bitEnterprise oMLX 20
Source artifacts
Raw machine-readable files for anyone who wants to dig deeper or run their own analysis.
- internal artifact memory/model-benchmark-reference.md