Eval Cogito V2 Preview 70B Nonreasoning: AI Model Benchmark Scores

Benchmark results for the cogito-70b-nonthinking model across 8 tasks, including AIME math problems and creative writing. The dataset was created by NousResearch and last updated on August 28, 2025. It includes scores for metrics like math_pass@1, overall_winrate, and drop_acc.

Use Cases

• Compare model performance on mathematical reasoning based on AIME24 and AIME25 benchmark scores.
• Evaluate language model capabilities in creative writing based on the creative-writing-v3 score.
• Assist in model selection for knowledge-intensive tasks using scores from GPQA Diamond and DROP benchmarks.
• Analyze model reasoning and factuality performance using the EQBench3 and ArenaHard evaluation results.

Strengths

• Provides scores across 8 distinct benchmarks, including AIME, ArenaHard, and DROP.
• Includes specific metric values, such as a 0.876 score on BBH and a 0.819 winrate on ArenaHard.
• Lists sample counts for each benchmark, such as 64 samples for AIME evaluations.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

NousResearch via Hugging Face.

Collection Method

Likely contains aggregated evaluation results from running the named model on standardized benchmarks.

Freshness

Last updated 2025-08-28 08:35:10; freshness should be verified.