it is indeed quite good!
don't try it in claude code/codex - those harnesses are overly tuned for their proprietary models
dcode (deepagents code) is a model agnostic harness - try it there with @FireworksAI_HQ :
``` dcode --model fireworks:accounts/fireworks/models/glm-5p2 ``` docs: https://docs.langchain.com/oss/python/deepagents/code/overview
Users praised LangChain CEO Harrison Chase's endorsement of GLM 5.2 paired with the new model-agnostic DeepAgents harness, calling the advice great.
No Digg Deeper questions have been answered for this story yet.
@hwchase17 @FireworksAI_HQ Great advice.
@janekm @FireworksAI_HQ anecdotally weve heard issues but maybe just our use case
Try it out! We are seeing amazing results with GLM 5.2!
it is indeed quite good!
don't try it in claude code/codex - those harnesses are overly tuned for their proprietary models
dcode (deepagents code) is a model agnostic harness - try it there with @FireworksAI_HQ :
``` dcode --model fireworks:accounts/fireworks/models/glm-5p2 ``` docs: https://docs.langchain.com/oss/python/deepagents/code/overview
@hwchase17 @FireworksAI_HQ It seems to work fairly well in Claude Code… probably they did train it on that harness
Deepagents code is sick cuz you can just use the best model as it comes out
it is indeed quite good!
don't try it in claude code/codex - those harnesses are overly tuned for their proprietary models
dcode (deepagents code) is a model agnostic harness - try it there with @FireworksAI_HQ :
``` dcode --model fireworks:accounts/fireworks/models/glm-5p2 ``` docs: https://docs.langchain.com/oss/python/deepagents/code/overview
@hwchase17 @FireworksAI_HQ To be fair I haven’t tried other harnesses… it might work better!
@hwchase17 @FireworksAI_HQ any speed comparison against deepseek v3 yet?
@pranavj09 when firework pass?
@hwchase17 @FireworksAI_HQ try zcode by zai
@hwchase17 @FireworksAI_HQ Model-agnostic is the right boundary. The harness should own task state, tool schemas, eval fixtures, and replay traces. Then swapping GLM 5.2, Claude, or Codex is an experiment instead of a new hidden runtime.
@hwchase17 @FireworksAI_HQ dcode being model-agnostic is actually a bigger deal than most realize—lets you actually A/B test inference cost vs. quality across providers without vendor lock-in. GLM-5p2 on Fireworks is an interesting pick for that.
