chat model, answer-letter probabilities · v1 set (earlier version)
Qwen3.8 Flash Next
Earlier version
Results f24d1f41, updated 16 November 2026, 22:20 UTC; scored by GNU Backgammon 1.08, 2-ply.
How we called it
- Model ID
- qwen3.8-flash-next
- Provider or endpoint
- our GPUs (vLLM, FP8, 4-way tensor parallel) behind Cygnet's letter-reading server
- Date run
- Missing
- Price
- no price per token; own hardware; price date not recorded
- Model type
- chat model, answer-letter probabilities
- Weights
- open
- Runs on
- our GPUs
- Reasoning setting
- Missing
- Confidence source
- letter probability
- Rules text
- short (the v1 rules text; the same stored request text for every model)
- Option groups
- not recorded; no decision in this data was split
- Sampling settings
- temperature 1.0
- Notes
- Qwen3.8-Flash-Next-FP8
Missing B7 fields: reasoning, option_groups, date_run
Results in v1 set (earlier version)
Earlier version: plays were listed in the runner's order, which tends to put the best play early. When Gemini 4 Argon answered the same 1,000 positions with the plays shuffled, its error rate changed by +1.9 (95% interval −0.2 to +4.2), so the effect is not proven.
- Error rate (PR)
- 31.127.3 to 35.2
- Decisions
- 1,000 (1,000 scored, 0 invalid)
- Engine's best play
- 47.3%
- Blunder rate
- 23.0%
- P95 equity loss
- 0.284
- Mean loss, worst 5%
- 0.488
- Median / P90 s per decision
- 0.46 / 0.71
- $ per 1,000 decisions
- Not measured
- Risk judgement, margin 0.001
- 61.6 (95% interval 58.6 to 64.7); risks found 37.5% (133 / 355), right safety 85.6% (316 / 369)
- Risk judgement, margin 0.02
- 60.9 (95% interval 57.9 to 64.1); risks found 37.7% (120 / 318), right safety 84.0% (341 / 406)
- No trade-off
- 276 (left out of every risk metric)
Risk call card
61.6risk judgement, margin 0.001
right risks133
wrong risks38
needless risks53
missed risks184
right safety316
Hit risk only: being hit. Gammon risk and priming are not in it.
By game phase
| Phase | Error rate (PR) | Decisions |
|---|---|---|
| Opening | Too few decisions | 143 |
| Contact | 37.9 | 641 |
| Holding game | Too few decisions | 20 |
| Race | Too few decisions | 136 |
| Bear-off | Too few decisions | 60 |