Wow.
@Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and not too verbose.
It responds with nuance and judgement, & handles long context VERY well.
I've never experienced an open weights model like this before.
@Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and not too verbose.
It responds with nuance and judgement, & handles long context VERY well.
I've never experienced an open weights model like this before.
Jun 18, 2026 · 11:52 PM UTC
230
484
7,289
891,433
The one big gap is that it is blind - it can't handle images at all.
If they fix that, then perhaps this could become the best available model in the world.
If they fix that, then perhaps this could become the best available model in the world.
19
11
540
58,716
BTW I've been using using @FireworksAI_HQ for inference. Other providers might not be as fast.
15
13
377
50,788
@jeremyphoward @Zai_org I don’t think they can easily add vision support to the model. Do they ?
4
0
3
16,469
@jeremyphoward @Zai_org no, it’s actually very slow and their servers are unreliable
4
0
35
24,221
@jeremyphoward @Zai_org Enjoy while it lasts, it will be banned soon. If it is not, it will destroy all the investment made in the AI buildout.
All chinese models will be banned. Even open sourced ones. How? Easy, there will be consecuences if US companies are found using them. Mark my words
All chinese models will be banned. Even open sourced ones. How? Easy, there will be consecuences if US companies are found using them. Mark my words
13
0
29
16,010
@jeremyphoward @Zai_org Any good in non agentic tasks? Like research or being a good mentor/pair programming help? I want it to help me solve problems, not solve them for me.
1
1
11
11,267
@KamStaszewski @Zai_org Yes really good at mentoring/pairing. It doesn't barge in and take over like GPT 5.5 tends to do.
0
0
30
9,642
@jeremyphoward @Zai_org I assume you're testing in your custom harness?
1
0
2
6,165
@jeremyphoward @Zai_org GLM-5.2 is between GPT-5.5 and Opus 4.8 in our new agentic knowledge work eval that we released just today. Very impressive
x.com/ArtificialAnly…
x.com/ArtificialAnly…
Announcing AA-Briefcase, the benchmark for the next era of agentic knowledge work
AA-Briefcase is our new benchmark for testing models on long-horizon knowledge work tasks in complex projects built by industry experts. Models are evaluated on multi-week projects, each with many nds of input source files.
We evaluated Claude Fable 5 from @AnthropicAI before it became unavailable, and it currently leads with an Elo score of 1587, followed by Claude Opus 4.8 (max, 1356), Opus 4.7, and the recently-released GLM 5.2 (max, 1266) from @Zai_org.
Claude Fable 5 cost $31 on average to run each AA-Briefcase task, followed by Claude Opus 4.8 at $10.40, GPT-5.5 (xhigh) at $3.68 and GLM-5.2 (max) at $2.40.
AA-Briefcase comprises four private scenarios, each representing a multi-week knowledge work project set in a realistic organizational context. A public fifth scenario has been released via @huggingface as a representation of scenario structure, submission, and grading (AA-Briefcase Lite). This does not count toward official AA-Briefcase results, and is demonstrative only.
Key elements of AA-Briefcase:
➤ Realistic long-horizon projects: AA-Briefcase moves beyond single, disconnected prompts by evaluating models across a coherent long-horizon project. Tasks build week by week, draw on shared institutional context, and require deliverables such as financial models, board presentations, and design mock-ups
➤ Large volumes of fragmented context: AA-Briefcase requires models to reason across thousands of inputs, including company documents, meeting transcripts, large-scale data exports, 25,000+ Slack messages and 3,500+ emails. These sources are fragmented, messy, and often contain realistic contradiction, testing whether models can navigate the ambiguity of real-world knowledge work
➤ Composite rubric and pairwise grading: AA-Briefcase combines binary rubric checks for ground-truth correctness with pairwise grading on analytical quality and presentation quality. Unlike many evaluations that focus on a single metric, AA-Briefcase tests agentic capabilities more comprehensively, exposing cases where models produce outputs that look polished but are incorrect or lack analytical rigor
➤ Built by industry experts: AA-Briefcase scenarios mirror real-world knowledge work, with tasks developed over months by experts across data science, product management and corporate strategy from companies including Google, McKinsey & Company and BCG. Task challenges are drawn from professional experience, making AA-Briefcase more reflective of the ambiguity, messy context and competing priorities that define real-world knowledge work
Key results:
➤ Claude Fable 5 leads AA-Briefcase at 1587 Elo: This is followed by Claude Opus 4.8 (1356) with the next-best non-Anthropic model, GLM-5.2 (max), ~90 points back at 1266. Note that Claude Fable 5 did not use the Opus 4.8 fallback for any task in AA-Briefcase
➤ Cost per task varies by ~800x across models tested: Claude Fable 5 leads the benchmark but costs more than $31 per task on average, compared to ~$0.04 for DeepSeek V4 Flash (max). The strongest price/performance options are open weights models such as GLM-5.2 (max) and DeepSeek V4 Pro (max), with GLM-5.2 (max) scoring only ~90 Elo below Claude Opus 4.8 (max) for less than 25% of the cost
➤ Real-world complexity remains difficult for models: The top performer, Claude Fable 5, satisfies all rubric criteria on just 3% of AA-Briefcase tasks. On 31 of 91 tasks, no model scores above 50% on the rubric criteria
➤ Task difficulty scales with the number of required input files: For each rubric check, we identify the set of source files needed to pass. Across all models, pass rates fall as this file count increases, though top-tier models degrade less than weaker models
More details below in thread ⬇️
4
13
233
22,158
@jeremyphoward @Zai_org Took me like 10 queries to suddenly start thinking of treating it more like opus 4.8 and 5.5. Within an hour i was committed to using it as a full replacement.
And then I spent way too much money on gpus to run it locally.
And then I spent way too much money on gpus to run it locally.
6
1
117
5,224
@jeremyphoward @Zai_org I've also found it to be extremely good, especially for UI tasks!
x.com/nutlope/status…
x.com/nutlope/status…
This model is insane at design.
I asked GLM 5.2 (left) and Opus 4.8 (right) to build me a landing page and you can't even tell the difference.
GLM cost $0.06 while opus cost $0.49. More than 6x cheaper while being faster + more token efficient.
Another win for open source AI.
0
1
30
4,960
@jeremyphoward @Zai_org GLM 5.2 has many good qualities but "super fast" is not one of them
3
0
17
4,092
@jeremyphowardi@ClementDelanguee@Zai_orgly use kimi-2.6 as a default on fireworks for some things on LangSmith and it’s great
which harness did you use? Harness eng for open models is def much less done than closed ones, we’re working on publishing some numbers around this, would love for the community to have a good set of harnesses for these frontier open models
which harness did you use? Harness eng for open models is def much less done than closed ones, we’re working on publishing some numbers around this, would love for the community to have a good set of harnesses for these frontier open models
0
0
16
1,504
@jeremyphoward @Zai_org It seems to need to generate a lot more reasoning tokens than the equivalent closed models. But otherwise it's shockingly good and feels very "complete".
0
0
15
2,044
@jeremyphoward @Zai_org Feels like about Opus 4.6-4.7 to me. Which is very impressive. Design skills are very good. I'll try some long context requests.
1
0
11
1,637
@jeremyphoward @Zai_org We agree! We even found it to be better than Opus 4.8, GPT 5.5, and all other models at frontend: x.com/designarena/st…
0
0
12
766
@jeremyphoward @Zai_org Had the same experience - this is a solid model
This is a watershed moment.
GLM-5.2 solidly beat Opus 4.8 and human participants in our backend take-home, making the whole thing obsolete.
It also pushed forward the state-of-the-art for multi-stage media-to-transcript, with a new release: offmute-v2.
I come with receipts.
0
0
5
848
@jeremyphoward @Zai_org High praise! The numbers look great but means even more from practitioners using it an actually enjoying the model.
0
0
5
257
