Throughput (tokens / sec)
avg. 71.8tokens/s
Long-horizon engineering with planning depth
Run instantly. Pay only for what you use.
GLM 5.1 is Z.ai's open-weight agentic engineering flagship, released under MIT: a 754B-parameter MoE with 40B active per token, a 200K-token context window, and up to 128K output tokens — room for entire generated modules in one response.
It is built for long-horizon engineering: sessions sustain up to 8 hours of autonomous planning, execution, and testing — in Z.ai's launch demo it built a working Linux desktop across 655 iterations without human input.
It tops the hardest coding leaderboard: #1 on SWE-bench Pro at 58.4, ahead of the closed frontier models behind it, with 70.6 on τ³-Bench and 71.8 on MCP-Atlas for tool-driven work.
#1on SWE-bench Pro
Score 58.4 — the top open model on the hardest coding leaderboard.
8hautonomous sessions
Planning, execution, and testing loops that run without human input.
128Kmax output tokens
Whole modules, reports, or test suites in a single response.
GLM 5.1 is strongest where engineering work is measured in hours, not requests — deep refactors, migration projects, and research agents that plan before they act.
58.4 on SWE-bench Pro — ahead of every closed frontier model on the contamination-resistant set.
Up to 8-hour runs — Z.ai's launch demo built a working Linux desktop across 655 planning-execution-test iterations.
70.6 on τ³-Bench and 71.8 on MCP-Atlas — tool selection that holds across long call chains.
68.0% on BrowseComp — multi-hop browsing and synthesis for research and analysis agents.
Long-horizon engineering runs
Plans, executes, and tests autonomously for up to 8 hours — refactors and migrations that outlast a workday.
The hardest software-engineering tasks
#1 on SWE-bench Pro (58.4) — the open model to beat on contamination-resistant coding.
Tool-driven agent pipelines
70.6 on τ³-Bench and 71.8 on MCP-Atlas — dependable tool selection across long call chains.
Very long generated outputs
A 128K-token output ceiling emits whole modules, reports, or test suites in a single response.
Cost-sensitive, high-volume fleets
Premium 40B-active pricing. For always-on loops where per-token cost dominates, use Qwen 3.6 35B-A3B.
Image, audio, or video inputs
GLM 5.1 is text-only. For agentic work grounded in screenshots, use Kimi K2.6.
Contexts beyond 200K tokens
For whole-repo prompts past the 200K window, DeepSeek V4 Flash offers a 1M-token context.
avg. 71.8tokens/s
avg. 720ms
58.4SWE-bench Pro
No. 1 on the contamination-resistant software-engineering leaderboard.
70.6τ³-Bench
Reliable tool use across long, multi-step agent interactions.
68.0%BrowseComp
Multi-hop web research and synthesis tasks.
Code examples
curl -X POST "https://api.entrim.ai/v1/chat/completions" \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $ENTRIM_API_KEY" \ -d '{ "model": "zai-org/GLM-5.1", "messages": [ { "role": "user", "content": "Why do they call it a building if it is already built?" } ] }'SDKs & docs
Use familiar SDK patterns with quickstart examples for common production setups.
OpenAI-compatible
Keep the request format your team already knows.
No cold starts
Production requests are served without model spin-up delays.
Marathon coding agents with visual context
High-throughput agentic work with 1M context
Dense agentic coding with thinking control