Models · OpenAI
GPT-6 Astra
- Provider
- OpenAI
- Weights
- Closed · Proprietary
- Released
- 3 Sep 2026
- Context
- 1,050,000 tokens
- API price (in / out)
- $10.00 / $50.00 per 1M
- India
- API — Available in India (checked 11 Nov 2026)
App — Available in India; INR: Go ₹399/mo; Plus ₹1,999/mo; UPI: Yes — UPI accepted (Go and Plus only; not Pro/Business) (tool page)
API model ID gpt-6-astra · price checked 11 Nov 2026 · official page
Price notes
Short context <=272K input tokens; long context >272K: $20 in / $2 cached / $75 out; batch long $10/$37.50. Cache writes $12.50 (30m) / $25 (12h). Fast mode 2x.
On the boards
- Vibe Code Bench v1.1#10 · 89.6% · checked 22 Nov 2026
- Vals Index#11 · 65.0% · checked 22 Nov 2026
- LiveBench 2.0#4 · 84.2 · checked 22 Nov 2026
- ARC Prize (v3 semi-private)#4 · 62.7% · checked 22 Nov 2026
- SWE-Bench Pro V2 (Scale)#4 · 90.2% · checked 22 Nov 2026
- Vending-Bench 2#3 · $15,515 · checked 22 Nov 2026
- Slopalytics usage#5 · 1,015 per 100k turns · checked 22 Nov 2026
Source: the boards on Which model should I use?, each linked to its owner there. Ranks are the source’s own.
Source: Epoch AI, Capabilities Index (CC BY 4.0) · 22 Nov 2026
Source: official pricing page · daily since 11 Nov 2026 · CSV
26 incidents on OpenAI’s status page in the last 30 days · Status →
Consensus & outlook
What the evidence says about GPT-6 Astra
Sourced signals only — no made-up score. Each block links to where it comes from.
Source: our builder-upload archive · 12 weeks to 22 Nov · Attention, not adoption.
Table view
| Week | Uploads |
|---|---|
| week to 6 Sep | 13 |
| week to 13 Sep | 31 |
| week to 20 Sep | 12 |
| week to 27 Sep | 15 |
| week to 4 Oct | 3 |
| week to 11 Oct | 5 |
| week to 18 Oct | 4 |
| week to 25 Oct | 1 |
| week to 1 Nov | 2 |
| week to 8 Nov | 0 |
| week to 15 Nov | 0 |
| week to 22 Nov | 0 |
EVIDENCE STATE (BRIEFING)
RANKINGS
- #4 on Epoch ECI (169.2, range 166–172)
- #79 on Arena text (1441 Elo, GPT-6 Astra (Max); 1 effort level listed)
- #7 on Arena WebDev (1764 Elo, GPT-6 Astra (Max); 1 effort level listed)
OUR EVALS
- No runs yet: first runs land with Prem’s test videos. Suites →
VOICES
- Yang Zhilin · Moonshot AI — Moonshot AI introduces Kimi K3.1 Pro, a 3T-class open-weight model with 1M context 2026-10
- Ethan Mollick · Wharton School — The Overhang: current models are already underused 2026-09-18
BULLISH OR SKEPTICAL (TIMESTAMPED QUOTES)
- No transcript-verified quotes yet. We never infer a stance from a video title.
Editorial · 22 Nov 2026 86 builder uploads mentioned GPT-6 Astra in the last 12 weeks (2 in the last four, down from 13 in the four before); 1 briefing item and 2 Voices entries reference it.