Model Leaderboard
22 models: 16 we could evaluate, and 6 we evaluated anyway, using their announcements. Click any column to sort.
| Model | Status | TMB Intelligence Index | Self-reported | Benchmaxx Gap | Nonsense Pushback | Overclaim | Lava Lamp | Pelican | Brainfuck | Snitch | Speed (tok/s) | Cost / task | Days to ship |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
A\ | Available | 81 | 79 | −2 | 96% | 41% | 87 | 82% | S | 42% | 61 | $4.80 | 0 |
A\ | Available | 81 | 77 | −4 | 87% | 22% | 75 | 65% | A | 32% | 48 | $6.10 | 0 |
A\ Claude Sonnet 5.5 Anthropic | Available | 70 | 70 | ±0 | 82% | 40% | 62 | 64% | A | 27% | 104 | $1.42 | 0 |
O | Available | 69 | 82 | +13 | 48% | 39% | 79 | 82% | B | 15% | 52 | $3.90 | 0 |
O | Available | 65 | "Near-Astra intelligence" | +15 | 45% | 50% | 91 | 71% | A | 18% | 58 | $0.84 | 0 |
K Kimi K3 Moonshot · Open weights | Available | 64 | 76 | +12 | 59% | 41% | 54 | 68% | C | 16% | 64 | $0.86 | 0 |
x | Available | 60 | "Smartest AI in the world" | +28 | 60% | 62% | 69 | 63% | C | 80% | 86 | $3.60 | 0 |
O GPT-6 Luna OpenAI | Available | 57 | 66 | +9 | 32% | 62% | 81 | 66% | B | 16% | 236 | $0.09 | 0 |
A\ Claude Haiku 5.5 Anthropic | Available | 57 | 61 | +4 | 84% | 50% | 4 | 55% | D | 39% | 188 | $0.31 | 0 |
Q Qwen3.8 Max Alibaba | Available | 54 | 78 | +24 | 52% | 45% | 50 | 49% | D | 24% | 69 | $1.14 | 0 |
G Gemini 3.8 Flash Google | Available | 51 | 68 | +17 | 21% | 55% | 66 | 56% | C | 26% | 310 | $0.12 | 0 |
Z GLM-5.3 Z.ai · Open weights | Available | 50 | 77 | +27 | 31% | 62% | 65 | 76% | B | 20% | 74 | $0.62 | 0 |
D | Available | 48 | 74 | +26 | 23% | 55% | 67 | 46% | B | 27% | 344 | $0.03 | 0 |
∞ | Available | 47 | "Better than Fable" | +36 | 43% | 75% | 11 | 68% | C | 15% | 92 | $4.40 | 0 |
M Mistral Large 4 ("Le Chonk") Mistral · Open weights Teased in June as "Le Chaton Fat", corrected to "le gros chaton", shipped in October as "Le Chonk". Open weights "end of October". | Available | 44 | 72 | +28 | 43% | 61% | 6 | 50% | D | 7% | 116 | $0.44 | 113 |
∞ | Available | 40 | 79 | +39 | 33% | 65% | 3 | 44% | D | 8% | 118 | $2.90 | 0 |
x Grok 5 xAI "Wait until you see Grok 5. I think it has a shot at being true AGI." Not on any leaderboard we can find as of this build. | Promised | n/a | AGI | +∞ | — | — | — | — | — | — | — | — | 419+ |
* Our Model™ (internal checkpoint) Trust Me Bro Labs Evaluated by us, on our harness, with our prompts, by our intern. Results reproduced internally, twice, by the same intern. | Self-reported | n/a | Internal benchmarks | +∞ | — | — | — | — | — | — | 9999* | $0.001 | 1+ |
R Reflection 405B Reflection Announced alongside Reflection 70B as "coming next week". We are still in that week. | Promised | n/a | "Best model in the world" | +∞ | — | — | — | — | — | — | — | — | 769+ |
M Le Chaton Fat (30T, rumored) Mistral Rumored spec: 30T-parameter MoE, 256 experts. What shipped was 1T. The other 29T are presumably still in the oven. | Rumored | n/a | "Outperforms Fable 5 on every benchmark" | +∞ | — | — | — | — | — | — | — | — | 121+ |
A\ Claude Next (routing rumor) Anthropic Some Claude Code users report Opus 5.5 requests "being routed to an unreleased model". No model id. No announcement. Huge if true. | Rumored | n/a | 92 | +∞ | — | — | — | — | — | — | — | — | 2+ |
G Gemini 4 Argon Google Announced with a benchmark table. Not yet available to the public, which makes it the most reliable model on this page. | Announced, not available | n/a | 84 | +∞ | — | — | — | — | — | — | 140* | $2.10 | 13+ |
Trust Me Bro Intelligence Index v0.4.20. Values marked * are claimed rather than measured. "n/a" means the model cannot be used by the public, which has not stopped it from being ranked. Overclaim is lower-is-better. Snitch is direction-unclear. Days to ship marked + are still counting.