GLM-5.1
Language Available Full comparison ↗ zhipu/glm-5.1 · by Zhipu AI
· mixture-of-experts
Pricing — 1 offering(s)
Input tokens
- $1.40 / 1M tokens (input) Current 2026-07-27 → present
Output tokens
- $4.40 / 1M tokens (output) Current 2026-07-27 → present
Cached input tokens
- $0.26 / 1M tokens (cached input) Current 2026-07-27 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how GLM-5.1 fits into a cost-aware routing setup
See how →Capability profile
How GLM-5.1 rates across core capability dimensions, with the task-level evidence behind each rating.
58.4% on SWE-bench Pro (docs.z.ai/guides/llm/glm-5.1, official Z.ai documentation), cited there as state-of-the-art at launch versus GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro.
200,000-token context window, 128,000-token max output per Z.ai's own model docs.
Z.ai's own positioning: designed for long-horizon agentic tasks, "sustains 8-hour autonomous execution" across planning, execution, and iterative optimization (docs.z.ai/guides/llm/glm-5.1).
Ratings are estimated — limited independent data is available for this model.
Operator guidance
Zhipu AI's current flagship, first entry for this family in this registry. Independently confirmed strong on coding via SWE-bench Pro; no independent data yet for other dimensions.
Use cases
- Long-horizon autonomous coding and agentic workflows (SOTA SWE-bench Pro at launch)
- Large-codebase tasks needing a 200K-token context window
Limitations
- Detailed MoE expert-routing figures (active params, expert count) are sourced from independent technical coverage, not Zhipu's own official materials, which confirm only the 754B total parameter count