OpenAI

Tier S · Ship-to-prod default.

GPT-6.2 Astra

S tier for demanding end-to-end coding and product work.

API pricing

USD per million tokens · OpenAI API Standard (≤272K input tokens)
Input
$8
Output
$40
Cached input
$0.4
Context window
1.05M
Pricing details and specifications

Try the active visual tests

Open the generated scene to test it, or compare that prompt with another model.

6 runs

Concourse Divide

Run date: 2026-11-18

Generation provenance: 2.1 · single-prompt

Superbench 2.1 brief · tool-free single-message authoring, no repair turn

Visible agent rerouting, collision avoidance, and recovery from blocked connectivity.

Ember Glider

Run date: 2026-11-17

Generation provenance: 2.1 · repository-run

Superbench 2.1 returning brief · original 1.x run protocol

Sunset gliding journey testing flight energy management, checkpoint flow, and atmospheric scene craft.

Lightkeeper's Atelier

Run date: 2026-11-18

Generation provenance: 2.1 · single-prompt

Superbench 2.1 brief · tool-free single-message authoring, no repair turn

Material differentiation, dynamic light, geometric shadows, and art direction.

Low-Poly Tower Defense

Run date: 2026-11-17

Generation provenance: 2.1 · repository-run

Superbench 2.1 returning brief · original 1.x run protocol

Diorama tower defense testing economy balance, wave design, placement rules, and combat readability.

Wrecklight Salvage

Run date: 2026-11-18

Generation provenance: 2.1 · single-prompt

Superbench 2.1 brief · tool-free single-message authoring, no repair turn

Three-phase game progression with persistent physical cargo and a visible late-stage consequence.

Editorial verdict

Where this model fits

Back to all benchmarks →

GPT-6.2 Astra takes the Astra slot in S tier. Its six Superbench 2.1 runs cover four frozen single-prompt briefs and two returning games built under the original tool-assisted protocol. Each artifact has its own validation notes.

Best for

  • Complex implementation across multiple files
  • Interactive prototypes and 3D experiences
  • Agent-led product work with review

Watch out

  • Each brief has one attempt at medium reasoning
  • The validation notes identify behavior still to be verified
Ranking rationale and evidence
  1. All six artifacts rendered in the release browser checks.
  2. In Concourse Divide, all 48 travelers arrived within 31.3 simulation seconds of reopening South in the matched browser review.
  3. S tier is an editorial placement for demanding coding and product work.

Evidence

2026-11-18

Superbash editorial model ranking

Takeaway: GPT-6.2 Astra is currently placed in Tier S.

The November 2026 editorial roster places GPT-6.2 Astra at rank 1.

Open source →
2026-11-18

GPT-6.2 Astra Superbench 2.1 run record

Takeaway: Six artifacts with frozen prompt hashes and per-run validation.

The run record documents both authoring protocols and the observed browser behavior.

Open source →

Local benchmark run

The GPT-6.2 Sol and Astra comparison

Superbash Learn sourceNovember 18, 2026Methodology and run evidence →
Sources, pricing details and methodology

Run identity

Model specifications

Pricing source · checked 2026-11-18 ↗
Status
Available in this Codex session
Codex model ID
gpt-6.2-astra
Run date
November 18, 2026
Access route
OpenAI API and Codex. Superbench generation used Codex CLI at medium reasoning. Listed API rates describe provider pricing.

Above 272K input tokens, input and cache rates double and output rates increase by 50% for the full request.

Benchmark sources

Both models received the same frozen briefs at medium reasoning in Codex. Four briefs used one prompt per fresh session. Ember Glider and Low-Poly Tower Defense used the original repository-run protocol, which permits tools and iteration.

GPT-6.2 Astra Superbench 2.1 run record

The validation notes list observed behavior and checks still open. Each brief has a single attempt per model. The collection has no aggregate score.

Evaluation settings

    Edition coverage

    Recorded means a published artifact, not a behavioral pass. Historical returning results retain their original protocol.

    Superbench 2.1: 6/6 recorded · 0 pending · 0 not run
    • 15 Ember Glider · repository-run · Recorded
    • 14 Low-Poly Tower Defense · repository-run · Recorded
    • 21 Skyhook Cargo · single-prompt · Recorded
    • 22 Concourse Divide · single-prompt · Recorded
    • 23 Lightkeeper's Atelier · single-prompt · Recorded
    • 24 Wrecklight Salvage · single-prompt · Recorded