GLM-5.2. is Now Available on Ajelix

Z.ai’s latest flagship model arrives on Ajelix with a solid 1M-token context window, substantially stronger coding performance, and a new open architecture built for the most demanding work.

Starting today, Ajelix users can select GLM-5.2 as their model of choice inside Ajelix Chat. Released on June 16, 2026, GLM-5.2 is the direct successor to GLM-5.1 (which has been replaced in Ajelix with this model).

“GLM-5.2 raises the bar again. The jump from 5.1 to 5.2 in coding and agentic benchmarks is significant, especially on Terminal-Bench, where the improvement is over 17 points. For our users who push Ajelix on demanding, multi-step workflows, this is the model to be on.” – Arturs, CTO at Ajelix

What Is GLM-5.2?

GLM-5.2 is a 753-billion-parameter Mixture-of-Experts model from Z.ai, released under the MIT open-source license with no regional restrictions. It builds on the GLM-5 architecture with three headline improvements:

• Solid 1M Context: is the first model in the GLM series to stably and reliably sustain a full 1 million-token context window, not just nominally support it. That means genuinely long documents, codebases, and conversation histories handled without degradation.
• Advanced Coding with Flexible Effort. Stronger coding capabilities are paired with multiple thinking effort levels, letting the model balance performance and latency depending on task complexity.

Benchmark Performance

GLM-5.2 delivers top-tier results across reasoning, coding, and agentic benchmarks. Here’s how it compares against leading models:

The numbers most directly relevant to Ajelix users are the coding and agentic categories. On Terminal-Bench 2.1, GLM-5.2 jumps from 63.5 to 81.0 — a 17.5-point improvement over its predecessor, reflecting dramatically better real-world terminal and autonomous task completion. On DeepSWE, it climbs from 18 to 46.2, more than doubling GLM-5.1’s score. SWE-Bench Pro moves from 58.4 to 62.1, keeping GLM-5.2 competitive with the best proprietary models on professional software engineering tasks.

Why We Added GLM-5.2 to Ajelix

Our model selection has always centered on three criteria: capability breadth, reasoning reliability, and open availability. GLM-5.2 clears all three.

The MIT license means no access restrictions — globally, for all Ajelix users. The 1M context is not just a headline number; the IndexShare architecture makes it operationally stable for genuinely long workflows. And the coding improvements translate directly to what our users do daily: building dashboards, writing formulas, generating reports, automating pipelines, and running multi-tool agentic sessions.

340,000+ professionals already made the switch to Ajelix Agents
From Excel automation to full business apps, Ajelix is the AI workspace built for work that actually needs to get done.

Try it

When to Choose GLM-5.2 in Ajelix

Not every task needs the most advanced model. But for these scenarios, GLM-5.2 is the right pick:

1. Long agentic workflows and multi-step automation. GLM-5.2 holds performance across extended, multi-turn sessions. Complex Ajelix workflows involving many tools, chained decisions, and iterative outputs benefit directly from this.
2. Advanced coding and technical tasks. For script generation, API integrations, formula logic, Excel file creation, data pipelines, and debugging, this is the strongest open model currently available on Ajelix.
3. Heavy document and repository work. Tasks that require reading across large codebases, dense multi-document analyses, or full repository generation are where the 1M context pays off.
4. Research and reasoning-heavy tasks. With an AIME 2026 score of 99.2 and GPQA-Diamond at 91.2, GLM-5.2 handles rigorous logical and scientific reasoning well. For research workflows, complex analysis, or tasks where accuracy on hard questions matters, it’s a strong choice.

How to Switch Models in Ajelix

To select GLM-5.2, open the model dropdown at the top of Ajelix Chat and choose it from the list. You can switch between available models at any time; each conversation can use a different model depending on what the task demands.

GLM-5.2 is available now for all Ajelix plan users.