ModelBench
Model catalog

Models

Every model in the catalog with source-linked pricing, context limits, provider availability, and published benchmark results.

Compare Export CSV
5 models
Structured output openrouter
Ranked by MMLU-Pro
Sakana: Sakana Namazu T
Reasoning Tools JSON 90.3

Japanese-specialized reasoning model based on Kimi K2.6 and tuned for Japanese language, culture, and business workflows

sakana/sakana-namazu 2026-08-03 262.144K context $0.95/M input $4/M output
6 providers
DeepSeek: DeepSeek V4 Pro T
Reasoning Tools JSON Open weights 87.5

Open MoE flagship with million-token context for coding and long agent runs

deepseek/deepseek-v4-pro 2026-04-24 1M context $0.66/M input $1.98/M output
71 providers
NVIDIA: Nemotron 3 Ultra 550B A55B T
Reasoning Tools Open weights 86.8

Largest Nemotron 3 model for maximum open-weight reasoning and agent accuracy

nvidia/nemotron-3-ultra-550b-a55b 2026-06-04 1M context $0.1/M input $0.1/M output
17 providers
DeepSeek: DeepSeek V4 Flash T
Reasoning Tools JSON Open weights 86.2

Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work

deepseek/deepseek-v4-flash 2026-04-24 1M context $0.15/M input $0.6/M output
68 providers
NVIDIA: Nemotron 3.5 Lightning 30B A3B T
Reasoning Tools JSON Open weights 81.9

Fast NVIDIA Nemotron MoE for reliable agentic tasks across enterprise workloads

nvidia/nemotron-3.5-lightning 2026-08-11 262.144K context $0.025/M input $0.12/M output
13 providers