← Back to directory
OpenAI
GPT-5.1
Released Nov 12, 2025 · knowledge cutoff 2024-10
- Status
- Superseded
- Location
- United States
- Modality
- Multimodal
- Context window
- 400K
- Max output
- 128K
- Speed
- —
- Price ($/MTok in / out)
- $1.25 / $10
- Cost per task (high effort)
- $0.27
- Open weights
- No
Benchmarks4 of 8 reported
GPQA Diamond
88.1%
SWE-bench Verified
76.3%
Humanity's Last Exam
26.5%
LMArena Elo
1465
MMLU-Pro
—
Terminal-Bench 2.1
—
AIME (latest)
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Spends thinking time in proportion to difficulty, so easy turns come back fast
- Warmer and more conversational than GPT-5 without sliding back into flattery
- Holds an instruction across a long session instead of drifting back to its defaults
Weaknesses
- Still misjudges ambiguous prompts, occasionally answering a hard question from the fast path
- The gains are in feel and pacing rather than capability — easier to work with, not smarter
Shipped as Instant and Thinking variants.
For developers
API model strings
Not researched
Licence
Proprietary — weights not released
Retirement
No retirement announced
Lineage
No recorded predecessor