Edison Advances
Benchmarks / BixBench3

Overall (20-task cohort)

overall

19 runs · 19 models · evaluated by Score.

# Model ↕ Variant ↕ Mode ↕ Average score ↓ Average cost (USD) ↕ Average output tokens ↕ Avg. dur ↕ Date ↕
1 claude-fable-5.1 (with fallback) — max 0.515 $65.07 163.9k 369.9m 2026-09-21
2 claude-opus-5.5 New — max 0.502 $42.33 224.6k 326.6m 2026-09-26
3 gpt-6-astra — max 0.492 $44.23 207.5k 432.1m 2026-09-07
4 grok-4.6 — max 0.488 $64.94 101.6k 379.6m 2026-09-09
5 gpt-5.6-sol — max 0.480 $33.88 141.6k 431.1m 2026-08-26
6 kimi-k3 — max 0.470 $15.34 195.7k 405.0m 2026-08-26
7 gpt-6-sol New — max 0.470 $13.46 217.6k 373.2m 2026-09-26
8 glm-5.2 — max 0.463 $27.01 234.5k 479.4m 2026-08-26
9 claude-opus-4.8 — max 0.460 $52.52 286.9k 431.5m 2026-08-26
10 gpt-5.5 — max 0.431 $25.17 73.8k 343.6m 2026-08-26
11 gemini-3.5-flash — max 0.421 $43.83 358.8k 382.1m 2026-08-26
12 claude-opus-5 — max 0.406 $41.17 206.9k 332.3m 2026-08-26
13 gpt-6-luna New — max 0.369 $1.18 413.2k 444.1m 2026-09-26
14 gpt-5.4-mini — max 0.364 $8.47 421.8k 401.3m 2026-08-26
15 gpt-5.4-nano — max 0.275 $3.39 552.7k 322.9m 2026-08-26
16 gemini-3.1-pro-preview — max 0.248 $129.14 87.5k 556.0m 2026-08-26
17 claude-sonnet-4.6 — max 0.153 $88.26 131.2k 1087.6m 2026-08-26
18 claude-haiku-4.5 — max 0.015 $2.20 69.9k 51.5m 2026-08-26
19 gemini-3.1-flash-lite — max 0.000 $0.35 17.5k 44.4m 2026-08-26

Click column headers to sort. Click mode chips to filter.