What's Changed

• Evals: correctly pass temperature/max_tokens when using Responses API by @Maratyszcza in https://github.com/openai/gpt-oss/pull/174
• Metal: move sampling to GPU by @Maratyszcza in https://github.com/openai/gpt-oss/pull/175
• Metal: benchmark generation of 100 tokens instead of 1 by @Maratyszcza in https://github.com/openai/gpt-oss/pull/178
• Metal: support generating multiple tokens at once by @Maratyszcza in https://github.com/openai/gpt-oss/pull/179
• Adding prefill benchmarking for metal backend by @ibahmed-oai in https://github.com/openai/gpt-oss/pull/181
• Metal: tune threadgroup sizes by @Maratyszcza in https://github.com/openai/gpt-oss/pull/180
• Metal: Adding optimized dense matmul kernel to optimize prefill perf by @ibahmed-oai in https://github.com/openai/gpt-oss/pull/183
• Metal: fused QKV projection (matmul+RoPE) kernel by @Maratyszcza in https://github.com/openai/gpt-oss/pull/184
• [Bugfix]Capture stderr for python tool with uv as backend by @wuhang2014 in https://github.com/openai/gpt-oss/pull/182

New Contributors

• @ibahmed-oai made their first contribution in https://github.com/openai/gpt-oss/pull/181
• @wuhang2014 made their first contribution in https://github.com/openai/gpt-oss/pull/182

Full Changelog: https://github.com/openai/gpt-oss/compare/v0.0.6...v0.0.7