Coding agents go multi-vendor: Grok, Opus 4.8, and Qwen3.7-Max crowd the field in a single week

Three new coding models shipped in 48 hours, all chasing Claude Code's lead. The cheap one comes from xAI, the polished one from Anthropic, and the benchmark-topper from Alibaba — and the choice now is less about capability than about workflow fit.

The week of May 28, 2026 made one thing clear: agentic coding is no longer a one-horse race, and the pricing floor just dropped again. xAI shipped grok-build-0.1 at $1/m input and $2/m output, Anthropic released Opus 4.8 with a cheaper fast mode, and Alibaba claimed Qwen3.7-Max beat ChatGPT and Gemini on coding benchmarks. For practitioners, the interesting question isn't which model wins a leaderboard — it's which one fits the way you actually work.

The xAI launch is the most disruptive on price. grok-build-0.1, the same model behind the Grok Build CLI, ships day-one in Cursor, Kilo Code, OpenCode, and OpenClaw, plus OpenRouter and Vercel AI Gateway. At $1/$2 per million tokens (as of May 29, 2026), it undercuts most of the agentic-coding pack while distributing through the editors developers already use. Anthropic is moving the other direction — toward workflow depth. Alex Albert framed Opus 4.8's cheaper fast mode as a behavior change, not a price change: fast mode for interactive editing, normal mode for longer async tasks. And xda-developers' write-up of the Code w/ Claude conference underscores the same point — Claude Code now spans CLI, IDE, and Desktop, and the productivity gain comes from matching the surface to the task instead of forcing everything through the terminal.

Meanwhile the field around them is filling in. Mistral rebranded Le Chat as Vibe, an agent-building surface, and Mistral Medium 3.5 joined Microsoft Copilot Studio's model lineup on May 28, 2026 with in-region data control. DeepSeek R1 remains the cheap reasoning option for teams who want chain-of-thought coding without OpenAI o1 pricing.

For builders shipping side projects, the calculus shifts toward grok-build-0.1 or DeepSeek R1 — cheap enough to run agentic loops without watching the meter. For teams already invested in Claude Code, Opus 4.8 plus fast mode is a workflow upgrade, not a switching decision. As Simon Willison pointed out, 2026 coding budgets were set before anyone knew how good these tools would get — the overrun is the signal, not the scandal.

The bottom line: Pick your coding model by where you spend your day — editor, CLI, or async queue — because price parity is closing and workflow fit is now the real differentiator.

Sources