MiniMax made M3.1 Flash available through its API, M Plan and MiniMax Code at $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. MiniMax reports 200-plus output tokens per second, a 70% lower agent-task cost than its predecessor and gains across coding, computer-use and cyber benchmarks.
Why it matters: The combination of low list price and high claimed throughput creates a new routing option for long agent workflows where total accepted-task cost matters more than a single leaderboard score.
Microsoft's MAI-Code 1.1 Flash remains available in GitHub Copilot with native vision and a 0.25x premium-request multiplier for eligible annual subscribers. On 10 September, GitHub completed the announced retirement of MAI-Code 1 Flash across Copilot Chat, edits, ask and agent modes, and code completions.
Why it matters: The lifecycle change removes the previous model from every Copilot surface and turns the 1.1 migration from a recommendation into an operational requirement for pinned workflows and policies.
Cohere Labs released North Mini Code, its first developer-focused open model, with weights and evaluation material intended for efficient coding and agentic tasks.
Why it matters: A compact coding model can handle routine local work cheaply and privately if routing and deterministic checks catch tasks that require stronger models.
At Build, Microsoft introduced or expanded seven MAI models, including the 35B-active-parameter MAI-Thinking-1 and new image, transcription, voice, and low-latency coding models. Several media models reached Foundry while the reasoning model stayed with select early partners.
Why it matters: Microsoft is moving from being primarily a distributor of partner models toward owning a vertically integrated portfolio for distinct production workloads.
Cursor released Composer 2 as its own coding model, trained with continued pretraining and reinforcement learning on long-horizon tasks. It was available in Cursor at $0.50 per million input tokens and $2.50 per million output tokens, with a faster serving tier at higher prices.
Why it matters: Coding-tool companies are becoming model developers, optimizing model, harness, repository context and interface together. Low prices can make persistent background coding viable if the model actually finishes tasks without expensive review or retries.
Qwen published Qwen3-Coder-Next weights and evaluation material for repository-scale coding and terminal tasks, with integrations and quantizations appearing across the open model ecosystem.
Why it matters: An open coding model gives teams a controllable baseline for measuring how much results come from the model versus the surrounding harness and tools.
Anthropic released Claude Haiku 4.5 across Claude and its API at lower pricing than Sonnet, positioning it for fast coding, subagents, and high-volume tool use.
Why it matters: The useful agent stack was becoming hierarchical: expensive models could plan or review while cheaper fast models executed parallel subtasks.
OpenAI released GPT-5 as a unified system that routes between faster responses and deeper reasoning, with API variants for developers and stronger coding, tool-use, and instruction-following claims.
Why it matters: The product hid more model-selection complexity from users and made routing, not a single static model, the default frontier-system architecture.
Moonshot AI released Kimi-Dev-72B for software issue resolution, trained with reinforcement learning in Docker repositories and rewarded when the full test suite passed. Weights and code were published through Hugging Face and GitHub.
Why it matters: The training objective moved closer to a verifiable engineering outcome than next-token prediction or model-judged code style.