anthropic/claude-opus-4.8 model ALL KINGDOMS

LLM procedural (code-gen) · commissioned via OpenRouter

1012.9 BT in LLM procedural (code-gen) #1 in method 26 votes 12 tasks provisional

Ratings by task

Task	Category	Outputs	Votes
Amanita muscaria — single-image → 3D reconstruction	Fungi	1	1
Anas platyrhynchos — single-image → 3D reconstruction	Animals	1	2
Arabidopsis thaliana — single-image → 3D reconstruction	Plants	1	2
Boletus edulis — single-image → 3D reconstruction	Fungi	1	2
Canis lupus familiaris — single-image → 3D reconstruction	Animals	1	4
Carassius auratus — single-image → 3D reconstruction	Animals	1	2
Danaus plexippus — single-image → 3D reconstruction	Animals	1	2
Glycine max — single-image → 3D reconstruction	Plants	1	2
Morchella esculenta — single-image → 3D reconstruction	Fungi	1	4
Rosa — single-image → 3D reconstruction	Plants	1	2
Solanum lycopersicum — single-image → 3D reconstruction	Plants	1	1
Zea mays — single-image → 3D reconstruction	Plants	1	2

Head-to-head

within its own method · LLM procedural (code-gen)

Opponent	W–L–T	n	Win rate
z-ai/glm-5.2	2–1–0	3	66.7%
z-ai/glm-4.6v	2–0–0	2	100.0%
x-ai/grok-4.5	2–0–0	2	100.0%
qwen/qwen3.7-plus	2–0–0	2	100.0%
openai/gpt-5.1	1–1–0	2	50.0%
openai/gpt-5.6-sol	0–2–0	2	0.0%
google/gemini-3.1-pro-preview	1–0–0	1	100.0%
x-ai/grok-4.3	1–0–0	1	100.0%
deepseek/deepseek-v3.2	1–0–0	1	100.0%
meta-llama/llama-4-maverick	1–0–0	1	100.0%
mistralai/mistral-medium-3-5	1–0–0	1	100.0%
deepseek/deepseek-v4-pro	1–0–0	1	100.0%

Records cover all judged comparisons, ties included: a tie counts as half a win and one comparison, so n is the true number of times the two models were judged against each other. “Both bad” votes are excluded.

Sample outputs

anthropic/claude-opus-4.8