OpenAI has released gpt-oss-120B and gpt-oss-20B as open-weight reasoning models under Apache 2.0 licensing. Together AI is announcing same-day availability of these models on its infrastructure, offering serverless and dedicated endpoints at $0.15/1M input tokens and $0.60/1M output tokens. The platform provides 99.9% uptime SLA, SOC 2 compliance, OpenAI-compatible APIs, fine-tuning support, and performance optimizations including FlashAttention and custom kernels. Benchmark results show gpt-oss-120B is competitive with OpenAI o3 on MMLU (90.0 vs 93.4) and AIME 2025 (97.9 vs 98.4).
