mirror of
https://github.com/hanzo-apps/hanzo.ai.git
synced 2026-10-28 07:32:52 +00:00
zandzeekay
6f8c6e2bdb
/ and /enso draw one chart again, the scatter d0199249 removed, read from the
runs when the page opens (lib/enso `field`): every single model at its best
priced run on the 198 questions, at the cost it filed; Enso Ultra, Enso and
Enso Flash on questions they have seen (194, 191 and 189 of 198), each at its
tier's cost on new questions; and the published scores of GPT-6 Sol, GPT-6
Luna, Kimi K3, GLM 5.3, Gemini 4 Argon and Grok 4.7, which no run has
measured, at their list rates times the median tokens a question took, a
dashed disc linking its source. A run of the same model replaces its
published score with no change to the code. The frontier is a line through
the points no other beats on both; a tier off it is not drawn.
The landing's title is a condition on the points: "Highest score. Cheapest
above 90%." while a tier tops every point and another is the cheapest above
90%, else "Accuracy against cost." The captions under the chart and over the
table are gone; each figure names its set where it stands. /enso says how
to route your own way: pinned, by policy, or trained on your own traffic.
66 lines
2.3 KiB
JSON
66 lines
2.3 KiB
JSON
{
|
|
"_README": "Published GPQA Diamond scores for models no enso.gpqa_diamond run has measured, drawn on the GPQA chart as reported points (lib/enso `field`). `id` is the api.hanzo.ai price-list id; a priced run of the same model on the whole set replaces its entry with no code change. `score` is the share right exactly as the source publishes it; `kind` is `lab` (the model's maker) or `independent` (an outside evaluation), the lab's figure preferred and Epoch AI's taken where the lab published none, one evaluator for all of them. `date` is the model's release: Epoch AI's pages state no evaluation date. Read 2026-10-17.",
|
|
"models": [
|
|
{
|
|
"id": "openai/gpt-6-sol",
|
|
"name": "GPT-6 Sol",
|
|
"score": 0.9431818181818182,
|
|
"setting": "gpt-6-sol_max",
|
|
"kind": "independent",
|
|
"publisher": "Epoch AI",
|
|
"source": "https://epoch.ai/models/gpt-6-sol",
|
|
"date": "2026-09-22"
|
|
},
|
|
{
|
|
"id": "openai/gpt-6-luna",
|
|
"name": "GPT-6 Luna",
|
|
"score": 0.9046717171717171,
|
|
"setting": "gpt-6-luna_max",
|
|
"kind": "independent",
|
|
"publisher": "Epoch AI",
|
|
"source": "https://epoch.ai/models/gpt-6-luna",
|
|
"date": "2026-09-22"
|
|
},
|
|
{
|
|
"id": "moonshotai/kimi-k3",
|
|
"name": "Kimi K3",
|
|
"score": 0.935,
|
|
"setting": "Kimi K3 (max): reasoning effort max, temperature 1.0, top-p 0.95, no tools",
|
|
"kind": "lab",
|
|
"publisher": "Moonshot AI",
|
|
"source": "https://github.com/MoonshotAI/Kimi-K3",
|
|
"date": "2026-07-16"
|
|
},
|
|
{
|
|
"id": "z-ai/glm-5.3",
|
|
"name": "GLM 5.3",
|
|
"score": 0.9090909090909091,
|
|
"setting": "glm-5.3_max",
|
|
"kind": "independent",
|
|
"publisher": "Epoch AI",
|
|
"source": "https://epoch.ai/models/glm-5-3",
|
|
"date": "2026-08-14"
|
|
},
|
|
{
|
|
"id": "google/gemini-4-argon",
|
|
"name": "Gemini 4 Argon",
|
|
"score": 0.9640151515151515,
|
|
"setting": "gemini-4-argon_high",
|
|
"kind": "independent",
|
|
"publisher": "Epoch AI",
|
|
"source": "https://epoch.ai/models/gemini-4-argon",
|
|
"date": "2026-09-30"
|
|
},
|
|
{
|
|
"id": "x-ai/grok-4.7",
|
|
"name": "Grok 4.7",
|
|
"score": 0.92739898989899,
|
|
"setting": "grok-4.7_xhigh",
|
|
"kind": "independent",
|
|
"publisher": "Epoch AI",
|
|
"source": "https://epoch.ai/models/grok-4-7",
|
|
"date": "2026-09-21"
|
|
}
|
|
]
|
|
}
|