Spaces:

---

smolagents

/

ml-intern

like 314

Running on CPU Upgrade

I am also experiencing the same issue. Below is a screenshot when attempting to use ML Intern

Here's the API response from POST https://smolagents-ml-intern.hf.space/api/session

{
    "detail": {
        "error": "premium_model_restricted",
        "message": "Premium models are gated to HF staff. Pick a free model — Kimi K2.6, MiniMax M2.7, GLM 5.1, or DeepSeek V4 Pro — instead."
    }
}

Same here but on cpu-basic.

A regression may have been introduced in the frontend_agent_config.json file for the model_name property.

It used to be set to huggingface/novita/moonshotai/kimi-k2.5, but now its set to ${ML_INTERN_CLAUDE_MODEL_ID:-bedrock/us.anthropic.claude-opus-4-6-v1}

Looks like a PR has already been merged to fix this issue: https://github.com/huggingface/ml-intern/pull/208/changes#diff-cab4b3771eaae4bdf5561838971323dd7169fd21d94d2522de0eb81fbde6fe38

And I'm in!

But now if we select opus 4.6 or gpt 5.5 it doesn't change, but it change to other models.

Its fixed now.

But now if we select opus 4.6 or gpt 5.5 it doesn't change, but it change to other models.

I am facing the same issue, only allowed to use kimi 2.6

But now if we select opus 4.6 or gpt 5.5 it doesn't change, but it change to other models.

I am facing the same issue, only allowed to use kimi 2.6

No I can use models other than opus 4.6 and gpt 5.5

Hi folks, yes this should be fixed now - let me know if not!

Still not working. Models part.

Can you solve this as still gpt 5.5 and opus 4.6 cannot not be selected.

Hi folks, yes this should be fixed now - let me know if not!

I am also not been able to access gpt 5.5 and opus 4.6

Can you try clearing the browser cache and trying again?

Tried after clearing cache of https://smolagents-ml-intern.hf.space/ but didn't work. Do I need to clear cache for huggingface.co too?

Ok this is very strange, let me try with a dummy account to see if I can reproduce it

Ok this is very strange, let me try with a dummy account to see if I can reproduce it

Any updates on this? @lewtun

Hi @automlmodel55 I implemented a fix on the OAuth which should force-clear cookies and prompt you to re-login (and hopefully solve the issue). Please let me know if it persists as I cannot replicate it (even with a dummy user)

@lewtun it is still the same problem when I tried with one of my friends account it still had same issue.

Friends account: @Meet1265

And for your information: I can't see any commits to file lately, as any updates to this space latest commit I can see on my side is from 12 hours earlier.

@lewtun After our discussion when I cloned repo and checked the code, I saw that claude opus 4.6 and gpt 5.5 are restricted as premium models and are explicitly marked gated only for hf staffs. But the error message:

"Premium models are gated to HF staff. Pick a free model — "
"Kimi K2.6, MiniMax M2.7, GLM 5.1, or DeepSeek V4 Pro — "
"instead."

on line 133 - 135 of backend/routes/agent.py is also not shown instead it provides an 403() Forbidden error in console.

I cross - checked my observations with chatgpt too and it spotted same problem of premium models being gated for hf users.

Can you solve it accordingly and can you explain why previously I was able to use that models but now it cannot be used freely. As previously I had some free message quota for these model and then after quota was exhausted it changed to free model. But now it starts with free models quota only.

@lewtun After our discussion when I cloned repo and checked the code, I saw that claude opus 4.6 and gpt 5.5 are restricted as premium models and are explicitly marked gated only for hf staffs. But the error message:

"Premium models are gated to HF staff. Pick a free model — "
"Kimi K2.6, MiniMax M2.7, GLM 5.1, or DeepSeek V4 Pro — "
"instead."

on line 133 - 135 of backend/routes/agent.py is also not shown instead it provides an 403() Forbidden error in console.

I cross - checked my observations with chatgpt too and it spotted same problem of premium models being gated for hf users.

Can you solve it accordingly and can you explain why previously I was able to use that models but now it cannot be used freely. As previously I had some free message quota for these model and then after quota was exhausted it changed to free model. But now it starts with free models quota only.

@lewtun This problem is solved. Thanks.