OpenAI · closed · 2026-04
GPT-5.5
#9 index 85.7
the takeOpenAI's current flagship, and it earns the top-tier billing on reasoning. The catch is the coding gap: OpenAI quotes 88.7 on SWE-bench, independent runs land near 80.6, and I trust the independent number. Still elite, still priced like it knows it.
SWE-bench Verified
80.6
GPQA Diamond
93.6
MMLU-Pro
—
- Context
- 1M
- Input
- $5/M
- Output
- $30/M
- Speed
- 79 tok/s
- Modality
- text, image
- Frontier reasoning (GPQA ~94 independently)
- Strong agentic coding
- 1M-token context
- $5/$30 is roughly double the prior tier
- Self-reported SWE-bench runs ~8 pts hot
- Long-context billing surcharge past 272K