DeepSeek-V4.1-Flash-1017 GGUF Benchmarks for 512K Context: Terminal Bench, DeepSWE.

Hey guys for those who are running DeepSeek locally, we ran agentic benchmarks including Terminal Bench 2.1, DeepSWE v1.1, Automation Bench v1.0.6, Aider Polyglot, and general benchmarks like SciCode, HLE Diamond (text-only), LiveCodeBench v6, GPQA Diamond, MMLU-Pro, AIME25, IFEval, MATH-500 (Level 5), HMMT 2025, HMMT 2026. We addressed some issues people mentioned with these new benchmarks. Surprisingly, the results are quite prominent with Unsloth quants performing >10% better on both DeepSWE and AutomationBench regardless of quant or size. Benchmarks are noisy - we did 35.1B tokens of benchmarking vs 1.3B before so 27x more data . See Benchmark runs per model where we did 64x AIME 2025 runs, 3x DeepSWE, 3x TB runs and more. 1-bit Unsloth GGUF scores 68.1 on DeepSWE v1.1, 72.2 on Terminal-Bench 2.1, 52.7 on AutomationBench v1.0.6. GGUF: huggingface.co/unsloth/DeepSeek-V4.1-Flash-1017-GGUF Full analysis: unsloth.ai/docs/models/deepseek-v4.1-flash