Model
Model explorer

GPT-6.1 Terra

CLOSED
OpenAI · GPT-6.1 family · released Oct 26, 2026

API id gpt-6.1-terra, released 2026-10-26 as 'a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing' and the successor to GPT-5.6 Terra. It costs $1/$5 per Mtok with cached input at $0.05 and 30-minute cache writes at $1.25, which OpenAI says is 50% below GPT-5.6 Terra and GPT-6.1 Sol. It has a 1.05M context with 922K max input. The launch post gives Agents' Last Exam, MMMU-Pro, OSWorld 2.1 and Terminal-Bench Science 0.1 only as charts and relative claims (e.g. Agents' Last Exam +15 pp over GPT-5.6 Terra at medium effort; Terminal-Bench Science within five points of Opus 5.5 at max), so none of those are filed. OpenAI treats it as Critical in cybersecurity under its Preparedness Framework and applies the GPT-6 Astra safeguards stack.

ReasoningCodingVisionFunction callingTool useAgentic
2997.6
Elo · rank #6
Parameters
Undisclosed
Active params
—
Context
1.05M tokens
Architecture
Proprietary model; architecture and parameter count undisclosed; 1.05M-token context (922K max input), text+image in / text out
License
Proprietary
Languages
—
API price (in/out)
$1 / $5
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
ExploitBenchAgents97.1%#3
best: GPT-6 Astra · 100.0%
HealthBench ProfessionalKnowledge62.7%#4
best: Claude Opus 5.5 · 65.6%
SEC-Bench ProCoding83.3%#1
best: this model · 83.3%
Terminal-Bench 4.0Agents49.5%#11
best: Claude Sonnet 5.5 · 70.6%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$1
Output / M tok
$5
GPT-6.1 family
Elo progression across releases
API price $1/$5 · each benchmark row carries its own source badge (see methodology)