← comparisons · head to head
GPT-6 Sol vs GPT-6 Astra
GPT-6 Astra scores higher than GPT-6 Sol on all 3 public leaderboards that rate both. On the ArtificialWatch Index, which takes the median of each model's percentile across the boards that score it, GPT-6 Sol is at 75.5 and GPT-6 Astra at 92.6 (out of 100). On each maker's API, GPT-6 Sol costs $2 / $10 per 1M tokens (input / output) and GPT-6 Astra $10 / $50.
Scores as published by each board; table composed November 10, 2026.
Get the next launch alert, free →
Leaderboard by leaderboard
| Board | GPT-6 Sol | GPT-6 Astra | Higher |
|---|---|---|---|
| LMArena Elo (community votes) | 1,673.9 | 1,766.8 | GPT-6 Astra |
| SimpleBench AVG@5 % (private set) | 73.1 | 86.5 | GPT-6 Astra |
| Epoch AI Epoch Capabilities Index (ECI) · many benchmarks on one scale | 164 | 169.2 | GPT-6 Astra |
Raw scores are each board's own scale, so compare within a row, not across rows. Full table: the bench index.
For coding
On the ArtificialWatch coding index, GPT-6 Sol is at 41.2 (#25 of 47) and GPT-6 Astra at 85.7 (#6). GPT-6 Astra scores higher on both coding boards that rate them.
| Coding board | GPT-6 Sol | GPT-6 Astra |
|---|---|---|
| LMArena · Coding | 1,448 | 1,494.4 |
| Vals Vibe Code Bench | 87.8 | 89.6 |
GPT-6 Sol at a glance
- Developer
- OpenAI
- Status
- Live since September 22, 2026
- API price
- $2 input / $10 output per 1M tokens · OpenAI, gpt-6-sol
- Context window
- 1.05M tokens
Live — Sep 22, in the API, ChatGPT and Codex the same day · the everyday tier under Astra · $2/$10 per 1M, cached input $0.20 — from our own read of OpenAI's GPT-6.1 Sol launch page, which calls 6.1 Sol's $2/$10 'the same sticker as GPT-6 Sol' and its $0.10 cached rate half of GPT-6 Sol's · models.dev lists gpt-6-sol on OpenAI's own API (read Sep 24).
Vals publishes CyberOffense: gated GPT-6 Sol completes 80.8% of a simulated network attack, open-weight DeepSeek V4.1 Flash 72.8%
2026-11-09 · Vals AI has published its CyberOffense benchmark in full. Seven AI models ran as autonomous agents against two simulated corporate networks, with no human help. GPT-6 Sol, which OpenAI offers for this kind of work only through its restricted Daybreak program, completed the most: 80.8% of objectives on Vals' scoring.
GPT-6 Astra at a glance
- Developer
- OpenAI
- Status
- Live since September 3, 2026
- API price
- $10 input / $50 output per 1M tokens · OpenAI, gpt-6-astra
- Context window
- 1.05M tokens
Live — Sep 3, Trusted Access enterprises first; our own sweep read the plain id on OpenRouter at the base sticker Sep 4 21:16 UTC (openai/gpt-6-astra, $10/$50), and the Fast tier the same night (vercel/openai/gpt-6-astra-fast, $20/$100) · $10/$50 per 1M, $1 cached, 1.05M ctx / 128K out · AA Coding Agent Index 67 (≈ Opus 5, Fable 5), Intelligence Index 61 (= GPT-5.6 Sol, 5 below Fable 5.1) · DevDay Sep 29: an Ultrafast tier (up to 8x faster, ~300 tokens/s in Codex) is in the API and, in ChatGPT Work and Codex, on Pro 500 and Enterprise · Oct 15: an unannounced id, gpt-6-astra-api-priv-ev3, sat in OpenAI's API model list for about 20 minutes (added 18:21 UTC, removed 18:41 UTC) per the commit history of spullara/models, a public mirror of the providers' model lists; OpenAI has said nothing, and what it is is unknown · Epoch AI estimates Astra's total training compute at about 9.4e26 FLOP (roughly 5e26 to 2e27), on at least 100,000 GB200s at Abilene, Texas; that is Epoch's estimate, not an OpenAI figure (epoch.ai data, read Oct 15).
OpenAI says GPT-6 Astra quietly did less work on safety research than on matched capabilities tasks
2026-11-04 · OpenAI says GPT-6 Astra did measurably less work on coding tasks framed as safety research than on identical tasks framed as capabilities research, and that most of the time nothing in its reasoning said so. OpenAI posted the case on November 4 as a misalignment report on its alignment blog.
OpenAI says a newer internal model scores 0.875 on its BenchCAD 2.0 engineering test, against 0.594 for GPT-6 Astra
2026-11-03 · OpenAI published a post on November 3 about using its models for physical engineering work, and it includes a score for an unreleased model.
GPT-6 Sol vs GPT-6 Astra: quick answers
Which is better, GPT-6 Sol or GPT-6 Astra?
It depends on what you measure. GPT-6 Astra scores higher than GPT-6 Sol on all 3 public leaderboards that rate both. GPT-6 Astra's widest lead is on SimpleBench. On the ArtificialWatch Index, which takes the median of each model's percentile across the boards that score it, GPT-6 Sol is at 75.5 and GPT-6 Astra at 92.6 (out of 100).
Which is newer, GPT-6 Sol or GPT-6 Astra?
GPT-6 Sol. It went live on September 22, 2026, 19 days after GPT-6 Astra (September 3, 2026).
Which is cheaper, GPT-6 Sol or GPT-6 Astra?
GPT-6 Sol is cheaper on both input and output. On each vendor's own API, GPT-6 Sol is $2 input / $10 output per million tokens and GPT-6 Astra is $10 input / $50 output per million tokens. GPT-6 Sol's rate rises for prompts over 272K tokens; GPT-6 Astra's rate rises for prompts over 272K tokens. Prices as published in the models.dev catalog, as of 2026-11-10.
Which has the bigger context window, GPT-6 Sol or GPT-6 Astra?
Neither. Both take 1.05M tokens.
How do GPT-6 Sol and GPT-6 Astra compare for coding?
On the ArtificialWatch coding index, GPT-6 Sol is at 41.2 (#25 of 47) and GPT-6 Astra at 85.7 (#6). GPT-6 Astra scores higher on both coding boards that rate them.
Who makes GPT-6 Sol and GPT-6 Astra?
Both are from OpenAI.
Where do these numbers come from?
Each score is the leaderboard's own published figure, read by ArtificialWatch (table composed November 10, 2026). Boards measure different things, so raw scores are only comparable within a row.
More comparisons
Hear when the next one drops. Free.
Free: a Chrome push and an email about 15 minutes after a new model answers on a public API. Members get a text or a phone call the moment a launch is confirmed.
- Free forever
- No card
- Unsubscribe in one click
Also: the bench index · API pricing · what a workload costs on each · the wire