Fireworks AI 推出了其 K2.7 Code 模型，这是其 K2 系列编码模型的又一进展，现已在其 serverless 平台和 API 上可用。该新模型旨在减少长代理循环中的推理 token 使用量，从而加快生成速度并降低每项已完成任务的成本。K2.7 Code 通过产生的推理 token 比其前身 K2.6 少约 30%，同时提高了编码基准测试的性能。 AI
影响 此次发布提高了代理编码任务的效率，有望降低 AI 开发者的运营成本。
排序理由 这是推理基础设施提供商的产品发布，而不是核心 AI 实验室的前沿模型发布。
在 X — Fireworks (inference infra) 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Available now on Fireworks serverless. → Standard tier (pay per token) → Priority tier for critical workloads → Fast path coming soon Pricing: $0.95 / 1M input, $4 / 1M output, $0.19 / 1M cache hits. 256K context. Full details here: https://t.co/R589xwqjf0 https://t.co/PwOW1Bjd…
In long agent loops, reasoning tokens get reused as context on every following turn. Shorter reasoning means smaller contexts downstream, faster generations, and fewer retries. K2.7 Code reduces that overhead without giving up quality, which lowers the real cost per completed
Moonshot released K2.7 Code, the latest in their K2 line of coding models, and it's live on Fireworks Day 0, on serverless and the API. It produces roughly 30% fewer reasoning tokens than K2.6 while scoring higher on Moonshot’s coding benchmarks. For agentic coding work, that h…