What's Changed

• Metal: add end-to-end benchmarks by @Maratyszcza in https://github.com/openai/gpt-oss/pull/161
• Metal: simplify and optimize Reponses API adapter by @Maratyszcza in https://github.com/openai/gpt-oss/pull/162
• Metal: fix KV-cache invalidation after reset+append by @Maratyszcza in https://github.com/openai/gpt-oss/pull/163
• Increase max output tokens in Reponses API to 131K by @Maratyszcza in https://github.com/openai/gpt-oss/pull/165
• Remove requirement on maximum Python version by @Maratyszcza in https://github.com/openai/gpt-oss/pull/167
• Move Lemonade to AMD section of awesome-gpt-oss by @danielholanda in https://github.com/openai/gpt-oss/pull/164
• Added VLLM Offline Serve working code. by @hrithiksagar-tih in https://github.com/openai/gpt-oss/pull/150
• Metal: indicate threadgroup is a multiple of simdgroup by @Maratyszcza in https://github.com/openai/gpt-oss/pull/168
• Metal: mlock model weights in memory by @Maratyszcza in https://github.com/openai/gpt-oss/pull/170
• Add You.com as tool for browser by @bojanbabic in https://github.com/openai/gpt-oss/pull/171

New Contributors

• @hrithiksagar-tih made their first contribution in https://github.com/openai/gpt-oss/pull/150
• @bojanbabic made their first contribution in https://github.com/openai/gpt-oss/pull/171

Full Changelog: https://github.com/openai/gpt-oss/compare/v0.0.5...v0.0.6