xtekky/gpt4free · error · ModelNotFoundError

LMArena Beta encountered an error: hasArenaError

Error message

LMArena Beta encountered an error: hasArenaError

What it means

While streaming the chat response from LMArena's Next.js flight protocol, a line prefixed 'a0:' carried the literal JSON string "hasArenaError". This is LMArena's own in-band error signal: the request reached the server and the stream started, but the arena backend rejected or failed the generation. g4f re-raises it as ModelNotFoundError.

Source

Thrown at g4f/Provider/needs_auth/LMArena.py:659

                data["modelBMessageId"] = modelBMessageId

            yield JsonRequest.from_dict(data)
            try:
                async with StreamSession(**args, timeout=timeout or 5 * 60) as session:
                    async with session.post(
                        url,
                        json=data,
                        proxy=proxy,
                    ) as response:
                        await raise_for_status(response)
                        args["cookies"] = merge_cookies(args["cookies"], response)
                        async for chunk in response.iter_lines():
                            line = chunk.decode()
                            yield PlainTextResponse(line)
                            if line.startswith("a0:"):
                                chunk = json.loads(line[3:])
                                if chunk == "hasArenaError":
                                    raise ModelNotFoundError(
                                        "LMArena Beta encountered an error: hasArenaError"
                                    )
                                yield chunk
                            elif line.startswith("b0:"):
                                ...
                            elif line.startswith("ag:"):
                                chunk = json.loads(line[3:])
                                yield Reasoning(chunk)
                            elif (
                                line.startswith("a2:") or line.startswith("b2:")
                            ) and line == 'a2:[{"type":"heartbeat"}]':
                                # 'a2:[{"type":"heartbeat"}]'
                                continue
                            elif line.startswith("a2:"):
                                chunk = json.loads(line[3:])
                                __images = [
                                    image.get("image")
                                    for image in chunk

View on GitHub (pinned to 973504e177)

Solutions

1. Retry with a different model from LMArena.get_models() — often only one arena model is affected
2. Retry the same request after a short wait; hasArenaError is frequently transient on the server side
3. Simplify the prompt/media payload if the error reproduces only for specific content
4. Check LMArena status/changelog and update g4f if the arena protocol changed

Example fix

// before
resp = await client.chat.completions.create(model='arena/model-x', messages=msgs)

// after
# rotate models on failure
for model in ['arena/model-x', 'arena/model-y']:
    try:
        resp = await client.chat.completions.create(model=model, messages=msgs); break
    except ModelNotFoundError: continue

from g4f.errors import ModelNotFoundError
for model in candidate_models:
    try:
        resp = await client.chat.completions.create(model=model, messages=msgs)
        break
    except ModelNotFoundError as e:
        if 'hasArenaError' in str(e):
            continue  # this arena model is failing server-side; try next
        raise

When it happens

Trigger: POSTing the conversation payload to LMArena succeeds, but during iter_lines() the model A stream ('a0:' prefix) yields the hasArenaError marker — e.g. when the selected model is temporarily disabled, the prompt triggers server-side moderation, or the arena service has an outage.

Common situations: A specific arena model being down or rate-limited server-side; prompts with content the arena refuses; using a model ID that exists in the catalog but is not currently servable; transient platform incidents.

Related errors

• LMArena: {json.loads(line[3:])}
• Invalid response: {last_msg}
• Failed to read response: {chunk.decode(errors='replace')}
• Ratelimit Exceeded!
• {chunk_text}

AI-assisted analysis of xtekky/gpt4free@973504e177 (2026-08-14). Data as JSON: /api/errors/16346a102966c2d7. Report an issue: GitHub.