Home  /  Models  /  GPT-6.2 Astra
OpenAI · GPT-6

GPT-6.2 Astra

GPT-6.2 Astra by OpenAI: League Score, pricing, context window and sourced benchmarks — the full tracked record.

97 Closed $8.00 / $40.00 per 1M 1.1M context

League positions: #2 overall · #1 Reasoning Division · #36 Value Division. Score coverage: 55% of components measured. Count tokens & cost for GPT-6.2 Astra · Estimate a workload cost

Agent-ready: this model is tracked with agentic strengths. See the AI Agents section for the products and frameworks that put models like this to work.

Specifications & pricing

MakerOpenAI — lab profile
FamilyGPT-6
ReleasedNov 17, 2026
StatusCurrent
WeightsClosed / proprietary API
LicenseProprietary
Knowledge cutoff2026-05-18
Context window1,050,000 tokens
Max output128,000 tokens
Modalities intext, image
Modalities outtext
API price (input)$8.00 per 1M tokens
API price (output)$40.00 per 1M tokens
Cached input$0.40 per 1M
Blended (3:1 in:out)$16.00 per 1M
Pricing sourceOfficial pricing page (checked Nov 18, 2026)
API accessOpenAI API

Benchmark results

BenchmarkScoreSourceDateLink
GPQA Diamond97.0 / 100BenchLM GPQA leaderboardNov 18, 2026Source
LiveBench overall84.9 / 100LiveBenchNov 18, 2026Source

How its League Score is built

ComponentRawNormalisedWeightDateLink
LiveBench overall84.995.540%Nov 18, 2026Source
GPQA Diamond97.0100.015%Nov 18, 2026Source

Normalised 0–100 against the tracked field; missing components redistribute their weight. Full rules on the methodology page.

Closest rivals

FAQ

How much does GPT-6.2 Astra cost?

GPT-6.2 Astra costs $8.00 per 1M input tokens and $40.00 per 1M output tokens on the first-party API (checked Nov 18, 2026).

What is the context window of GPT-6.2 Astra?

GPT-6.2 Astra accepts up to 1,050,000 tokens of context.

Is GPT-6.2 Astra open source?

No — GPT-6.2 Astra is a closed model available through OpenAI’s API and products.

Data verified Nov 18, 2026. Spotted an error? Every figure above links to its source.

From the blog: this model was released on Nov 17, 2026 and entered the League at #1 with a score of 100.0 from two field maxima — a LiveBench 84.9 and a GPQA Diamond 97.0 — at 55% coverage, tying Claude Fable 5.2, whose 100.0 rests on one component — see the GPT-6.2 Astra release analysis for both normalisations worked in full, the workload costing of the tie, and what the #1 ranking does not yet prove. On Nov 20, 2026 Gemini 4 Neon’s LiveBench 87.4 took the #1 spot at 100.0 on one component, and this model’s deeper record moved to #2 at 96.7 — see the Gemini 4 Neon first-score analysis for the two records compared.