Understand the numbers

The value frontier

Every model, plotted by what it costs against how good it is. The staircase is the frontier: for those models, no other model scores at least as high, costs no more, and wins on one of the two. Everything below and to the right has such a model on the staircase, though not always a like-for-like replacement.

Changing the workload below moves the frontier less than you would expect: 46% of frontier models are on it under every workload (7 move). What does change is the price — the same model swings up to 3.6× across these shapes. So the control is not about which models to consider. It is about what you would actually pay for them.

Workload
Measured on

General · Balanced

Claude Opus 5.5

Anthropic

on the frontier
General capability
1511.7
Effective $/M
$8.00/M

Compared with the previous price rung, Gemini 3.8 Flash: +14.7 Elo · +$5.00/M for Balanced. Score differences are not task-success or capability guarantees.

Compare General scores and capabilities
$0.030$0.100$0.300$1.00$3.00$10.00$30.0010801170126013501440effective $ per million tokens — log scaleGeneral capability
effective $ per million tokens — log scale · General capability

On the frontier

General · Balanced · Effective $/M

  1. 1 Llama 3.1 8B Instruct Meta 1186.5 $0.025/M cheapest rung
  2. 2 gpt-oss-20b OpenAI 1287.7 $0.036/M +101.2 Elo · +$0.011/M vs previous rung
  3. 3 Gemma 3 4B Google 1290.8 $0.063/M +3.1 Elo · +$0.027/M vs previous rung
  4. 4 gpt-oss-120b OpenAI 1365.4 $0.065/M +74.6 Elo · +$0.0025/M vs previous rung
  5. 5 DeepSeek V4 Flash 0423 DeepSeek 1432.1 $0.113/M +66.7 Elo · +$0.048/M vs previous rung
  6. 6 Gemma 4 26B A4B Google 1433.9 $0.127/M +1.8 Elo · +$0.014/M vs previous rung
  7. 7 MiMo-V2.6-Flash Xiaomi 1456.4 $0.175/M +22.5 Elo · +$0.048/M vs previous rung
  8. 8 DeepSeek V4.1 Flash DeepSeek 1462.5 $0.182/M +6.1 Elo · +$0.0073/M vs previous rung
  9. 9 GLM 5.3 Flash Z.ai 1469.6 $0.238/M +7.1 Elo · +$0.055/M vs previous rung
  10. 10 MiMo-V2.6-Pro Xiaomi 1491.0 $0.540/M +21.4 Elo · +$0.303/M vs previous rung
  11. 11 Gemini 3.8 Flash Google 1497.0 $3.00/M +6.0 Elo · +$2.46/M vs previous rung
  12. 12 Claude Opus 5.5 Anthropic 1511.7 $8.00/M +14.7 Elo · +$5.00/M vs previous rung

Cite this claim

Plain text and BibTeX
Plain
133 of 145 rated, priced models have an alternative with an equal-or-higher measured score and an equal-or-lower price, with a strict advantage on at least one axis. 12 are not dominated on these two axes. As of 2026-10-06. Lens: LMArena Elo (CC BY 4.0). Ranking source digest: 06cd12db54e2968c. https://undominated.ai/frontier/
Publication: sha256:ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932; calculation SHA-256: 5ae4d0bee9b70298dd437b05df87c32acb407040b50bf88240677f295f25955d; permitted inputs: https://undominated.ai/data/publications/manifests/ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932.json
BibTeX
@misc{undominated-dominated-census-2026-10-06,
  author = {{Undominated.ai}},
  title  = {Dominated census — Undominated.ai},
  year   = {2026},
  url    = {https://undominated.ai/frontier/},
  note   = {as of 2026-10-06; LMArena Elo (CC BY 4.0); ranking source 06cd12db54e2968c},
  annote = {Publication: sha256:ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932; calculation SHA-256: 5ae4d0bee9b70298dd437b05df87c32acb407040b50bf88240677f295f25955d; permitted inputs: https://undominated.ai/data/publications/manifests/ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932.json}
}

LMArena scores used under CC BY 4.0 · numbers recomputed from the catalogue on this page. Input identity & calculation SHA.

Evidence & Ask