chat model, one letter · League 3
GPT-6 Astramedium
FinishedPublished
Results f24d1f41, updated 16 November 2026, 21:24 UTC; scored by GNU Backgammon 1.08, 2-ply.
How we called it
- Model ID
- gpt-6-astra
- Provider or endpoint
- ChatGPT subscription
- Date run
- 2026-10-20 to 2026-10-24
- Price
- $10.0 / $50.0 per million tokens in / out; API-equivalent (subscription); price date not recorded
- Model type
- chat model, one letter
- Weights
- closed
- Runs on
- subscription
- Reasoning setting
- reasoning effort medium, the model's default
- Confidence source
- none
- Rules text
- long
- Option groups
- up to 26 options per request (letters A-Z); more legal plays are split into equal groups, a winner per group, then among the winners; in this data 182 decisions were split into groups of 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26
- Sampling settings
- not sent (API default)
- Notes
- ChatGPT subscription (Codex backend)
- Earlier name
- GPT-6 Astra (on) (renamed on 28 Oct 2026 to the reasoning effort its requests sent)
Missing B7 fields: price.date
Results in League 3
- Error rate (PR)
- 11.19.5 to 12.9
- Decisions
- 1,362 (1,362 scored, 0 invalid)
- Engine's best play
- 65.2%
- Blunder rate
- 8.6%
- P95 equity loss
- 0.125
- Mean loss, worst 5%
- 0.222
- Median / P90 s per decision
- 4.40 / 15.1
- $ per 1,000 decisions
- 50.268API-equivalent (subscription)
- Risk judgement, margin 0.001
- 74.7 (95% interval 71.9 to 77.3); risks found 54.9% (245 / 446), right safety 94.4% (542 / 574)
- Risk judgement, margin 0.02
- 74.9 (95% interval 72.4 to 77.3); risks found 56.5% (225 / 398), right safety 93.2% (580 / 622)
- No trade-off
- 342 (left out of every risk metric)
Risk call card
74.7risk judgement, margin 0.001
right risks245
wrong risks12
needless risks32
missed risks189
right safety542
Hit risk only: being hit. Gammon risk and priming are not in it.
Close choices
A decision is a close choice when the second most likely play has at least half the probability of the first (internally: split).
Probabilities over plays, not sampled reasoning. Method
- Decisions
- 1,362
- Eligible (full probability distribution)
- 0
- Left out, by reason
- one letter only, or a program 1,362
- Close choices
- N/A no full probability distribution
By game phase
| Phase | Error rate (PR) | Decisions |
|---|---|---|
| Opening | Too few decisions | 184 |
| Contact | 14.1 | 887 |
| Holding game | Too few decisions | 62 |
| Race | Too few decisions | 137 |
| Bear-off | Too few decisions | 92 |