GLM-5.1 access for coders
- BytePlus ModelArk announced paywalled access to GLM-5.1 for agentic coding at roughly $10 per month. - The offering claims top-tier performance on coding benchmarks and open-weight access without throttling. - This changes cost and availability assumptions for teams building local agentic dev tooling and CI integrations (x.com).
BytePlus has started selling paywalled access to GLM-5.1 on its ModelArk service, offering agentic coding under a subscription tier that begins around $10/month. (byteplus.com) BytePlus’s ModelArk “Coding Plan” lists GLM-5.1 among supported models and advertises tiered Lite/Pro plans and tool integrations. (byteplus.com) The company bundles GLM-5.1 with coding tools like Claude Code, Cursor, Cline, OpenClaw and Hermes Agent and promises “faster, more stable performance — without speed reductions.” (byteplus.com) GLM-5.1 itself is described by model pages as a 744‑billion-parameter mixture‑of‑experts model that activates ~40B parameters, supports a ~200k token context window, and posts leading scores on coding benchmarks such as SWE‑Bench Pro. (unsloth.ai) Running GLM-5.1 locally is nontrivial: published guides list a full‑model disk footprint around 1.65 TB and recommend H100/H200 or equivalent accelerators, though the model’s 40B active parameter count reduces continuous GPU requirements compared with serving all 744B. (unsloth.ai) The new ModelArk pricing sits below many paid coding subscriptions: Anthropic’s Claude Code Pro starts at about $20/month, while Max and team tiers run substantially higher, making BytePlus’s ~$10 entry price a lower‑cost option for developers. (claude.com) BytePlus updated its ModelArk pricing documentation around April 20, 2026, and GLM‑5.1 model pages were published in early April 2026, giving teams concrete dates to compare subscription availability versus self‑hosting timelines. (docs.byteplus.com)
