BOARD HISTORY · COMPARE EDITIONS
What changed?
v1.1 · rev 2 (September 30, 2026) → v1.1 · rev 4 (October 14, 2026) · the current board.
Scores comparable. Same board version (v1.1): score changes are directly comparable. Rank depends on the field: a model can move without its own score changing.
MODELS40 → 42on each board
NEW2added in the newer edition
NO LONGER LISTED0not a failure
RANK UP / DOWN0 / 39of 40 on both
ABC TEST CHANGES0PASS / FAIL / N/A verdicts
| Model | Rank | Change | Blind Bantu mean | Δ | ABC Test |
|---|---|---|---|---|---|
| Gemini 3.6 FlashGoogle | 1 → 1 | — | 16.8 → 16.8 | +0.0 | FAIL |
| Gemini 4 ArgonGoogle | — → 2 | New | 15.1 | — | FAIL |
| Claude Fable 5 ‡Anthropic | 2 → 3 | ▼ 1 | 14.7 → 14.7 | +0.0 | FAIL |
| Gemini 3.5 FlashGoogle | 3 → 4 | ▼ 1 | 14.6 → 14.6 | +0.0 | FAIL |
| Claude Fable 5.1Anthropic | 4 → 5 | ▼ 1 | 14.2 → 14.2 | +0.0 | FAIL |
| Claude Sonnet 5.5Anthropic | 5 → 6 | ▼ 1 | 14.0 → 14.0 | +0.0 | FAIL |
| Gemini 3 ◊Google | 6 → 7 | ▼ 1 | 13.9 → 13.9 | +0.0 | FAIL |
| Claude Opus 5.5Anthropic | 7 → 8 | ▼ 1 | 13.4 → 13.4 | +0.0 | FAIL |
| Gemini 3.7 FlashGoogle | 8 → 9 | ▼ 1 | 12.9 → 12.9 | +0.0 | FAIL |
| Qwen 3.8-2.4T-A95BAlibaba | 9 → 10 | ▼ 1 | 12.5 → 12.5 | +0.0 | FAIL |
| Qwen 3.8 MaxAlibaba | 10 → 11 | ▼ 1 | 12.0 → 12.0 | +0.0 | FAIL |
| GPT-5.5OpenAI | 11 → 12 | ▼ 1 | 11.8 → 11.8 | +0.0 | FAIL |
| Gemini 3.8 FlashGoogle | 12 → 13 | ▼ 1 | 11.7 → 11.7 | +0.0 | FAIL |
| DeepSeek V4 Pro 0813DeepSeek | 13 → 14 | ▼ 1 | 11.6 → 11.6 | +0.0 | FAIL |
| Grok 4.6SpaceXAI | 14 → 15 | ▼ 1 | 11.2 → 11.2 | +0.0 | FAIL |
| GLM-5.2Z.ai / Zhipu | 15 → 16 | ▼ 1 | 10.1 → 10.1 | +0.0 | FAIL |
| Claude Opus 4.8Anthropic | 16 → 17 | ▼ 1 | 9.4 → 9.4 | +0.0 | FAIL |
| Gemini 3.5 Flash-LiteGoogle | 17 → 18 | ▼ 1 | 9.1 → 9.1 | +0.0 | FAIL |
| DeepSeek V4DeepSeek | 18 → 19 | ▼ 1 | 9.0 → 9.0 | +0.0 | FAIL |
| DeepSeek V4 FlashDeepSeek | 19 → 20 | ▼ 1 | 8.9 → 8.9 | +0.0 | FAIL |
| Grok 4.5SpaceXAI | 20 → 21 | ▼ 1 | 8.7 → 8.7 | +0.0 | FAIL |
| GPT-5.6 TerraOpenAI | 21 → 22 | ▼ 1 | 8.1 → 8.1 | +0.0 | FAIL |
| GPT-6 AstraOpenAI | 22 → 23 | ▼ 1 | 7.4 → 7.4 | +0.0 | FAIL |
| Claude Haiku 5.5Anthropic | — → 24 | New | 7.1 | — | FAIL |
| Mistral Large 3Mistral AI | 23 → 25 | ▼ 2 | 7.0 → 7.0 | +0.0 | FAIL |
| Kimi K3Moonshot AI | 24 → 26 | ▼ 2 | 6.8 → 6.8 | +0.0 | FAIL |
| Qwen 3.7Alibaba | 25 → 27 | ▼ 2 | 6.7 → 6.7 | +0.0 | FAIL |
| GPT-5.6 SolOpenAI | 26 → 28 | ▼ 2 | 6.7 → 6.7 | +0.0 | FAIL |
| Nemotron 3 UltraNVIDIA | 27 → 29 | ▼ 2 | 6.2 → 6.2 | +0.0 | FAIL |
| MiniMax M3MiniMax | 28 → 30 | ▼ 2 | 5.7 → 5.7 | +0.0 | FAIL |
| GPT-5.6 LunaOpenAI | 29 → 31 | ▼ 2 | 5.2 → 5.2 | +0.0 | FAIL |
| Claude Sonnet 5Anthropic | 30 → 32 | ▼ 2 | 5.0 → 5.0 | +0.0 | FAIL |
| Kimi K2.7Moonshot AI | 31 → 33 | ▼ 2 | 4.9 → 4.9 | +0.0 | FAIL |
| GPT-OSS 120BOpenAI | 32 → 34 | ▼ 2 | 3.9 → 3.9 | +0.0 | FAIL |
| Command A+Cohere | 33 → 35 | ▼ 2 | 3.4 → 3.4 | +0.0 | FAIL |
| GLM-5.3 FlashZ.ai / Zhipu | 34 → 36 | ▼ 2 | 3.1 → 3.1 | +0.0 | FAIL |
| TInklingThinking Machines Lab | 35 → 37 | ▼ 2 | 3.0 → 3.0 | +0.0 | FAIL |
| Muse Glimmer 30BMeta | 36 → 38 | ▼ 2 | 2.0 → 2.0 | +0.0 | FAIL |
| Nemotron Lightning 3.5 30BNVIDIA | 37 → 39 | ▼ 2 | 1.5 → 1.5 | +0.0 | FAIL |
| Phi-4-reasoningMicrosoft | 38 → 40 | ▼ 2 | 0.2 → 0.2 | +0.0 | FAIL |
| Grok 4.7SpaceXAI | 39 → 41 | ▼ 2 | 0.0 → 0.0 | +0.0 | FAIL |
| GLM-5.3Z.ai / Zhipu | 40 → 42 | ▼ 2 | — → — | — | FAIL |
A model missing from an edition was not on that board (not yet evaluated, superseded or retired). Absence is never a benchmark failure. ABC Test verdicts (PASS / FAIL / N_A) are derived per edition from its own published scores and protocol; mastery_changes counts changed verdicts. First place is not a PASS.