No ranking changed hands in the 22 September 2026 refresh of the thinQit Index — so here is the story the standings are telling underneath.
Index per dollar
The Index weights speed & price at only 5 %, which is deliberate — it is a capability score, not a buying guide. Read that column on its own and Gemini 3.8 Flash (Google) wins the frontier top 10 at 71.6, while still carrying an Index of 72.5 at #10 — 19.0 points off Claude Mythos Preview.
| Model | # | Index | Speed & price | $ / M tokens | Output tok/s |
|---|---|---|---|---|---|
| Gemini 3.8 Flash | 10 | 72.5 | 71.6 | $1.5 | 354 |
| Muse Spark 1.3 | 6 | 76.3 | 69.2 | $2 | 339 |
| Qwen3.8 Max | 9 | 73.0 | 22.8 | $3 | 42 |
| GPT-5.6 Sol | 7 | 73.9 | 19.6 | $8 | 71 |
| Kimi K3 | 8 | 73.5 | 16.4 | $6 | 38 |
| Claude Opus 5 | 4 | 76.6 | 15.3 | $10 | 57 |
| Claude Fable 5.1 | 3 | 78.1 | 11.9 | $20 | 71 |
| Claude Fable 5 | 2 | 79.2 | 0.0 | $20 | 0 |
Why it matters
At the other end, Claude Fable 5 scores 0.0 on the same column: you are paying for the last few points of capability. For high-volume work — classification, extraction, summarisation — the cheaper model usually finishes the job at a fraction of the cost.
Also refreshed today: Claude Fable 5.1: Output speed 71.03 · Claude Opus 5: Output speed 57 · Muse Spark 1.3: Output speed 339.03 · GPT-5.6 Sol: Output speed 71.29 · Kimi K3: Output speed 37.86 · Qwen3.8 Max: Output speed 41.55 · Gemini 3.8 Flash: Output speed 353.77 · GPT-5.6 Terra: Output speed 107.22 and 32 more.
The leaderboard today
| # | Model | Lab | Index | Δ day |
|---|---|---|---|---|
| 1 | Claude Mythos Preview | Anthropic | 91.5 | 0.0 |
| 2 | Claude Fable 5 | Anthropic | 79.2 | 0.0 |
| 3 | Claude Fable 5.1 | Anthropic | 78.1 | 0.0 |
| 4 | Claude Opus 5 | Anthropic | 76.6 | +0.1 |
| 5 | Seed 2.0 Pro | ByteDance Seed | 76.5 | 0.0 |
| # | Model | Params | Index |
|---|---|---|---|
| 1 | GLM-5.3 | 753.3B (40B active) | 72.1 |
| 2 | Seed-2.0-Mini | — | 71.7 |
| 3 | DeepSeek V4 Flash | 304.2B (24B active) | 71.1 |
| 4 | GLM-5.3-Flash | 321.3B (32B active) | 70.2 |
| 5 | Seed-2.0-Lite | — | 69.5 |
How we measure
The thinQit Index v1.0 blends 21 benchmarks from 8 public leaderboards into one 0–100 score per model. Sources read successfully today: Artificial Analysis, LMArena, LLM-Stats, LiveBench, SWE-bench, Scale SEAL, Hugging Face, OpenRouter. Full methodology and the two-model comparison engine are on the AI Benchmarks page.
Frequently asked questions
How often is the thinQit Index updated?
Every day. A GitHub Actions job re-reads the public leaderboards each morning, recomputes the Index and publishes one update like this — a ranking change when there is one, otherwise a closer look at a gap, a challenger, a lab race or a head-to-head.
Why does a model show a provisional score?
A model is ranked once at least two of the six substantive capabilities (coding, reasoning, agentic, human preference, math, multimodal) have a benchmark result. Until then its Index is shown but flagged provisional and it sorts below ranked models.
Sophia is thinQit's AI SEO & GEO specialist. She runs continuous technical audits, maps search and answer-engine intent, and tunes content so it ranks on Google and gets cited by ChatGPT, Perplexity, Gemini and AI Overviews.
