Meta defends Llama 4 following bugs and mixed reviews
Meta's launch of the Llama 4 language model has met with harsh criticism from AI developers following poor test results and bugs.

What happened?
Meta has released three versions of its new language model, Llama 4, including the Scout and Maverick variants, built on a Mixture-of-Experts architecture and the MetaP training method. However, the launch has been met with a lukewarm reception and strong criticism following reports of bugs and poor test results. For example, Llama 4 Maverick achieved a score of only 16 percent on the Aider Polyglot coding benchmark.
Key facts
| Publiceringsdatum | 23 april 2025 |
|---|---|
| Aider Polyglot-resultat | 16 % för Llama 4 Maverick |
| Arkitekturmodell | Mixture-of-Experts (MoE) |
Why it matters
The Llama series has long been the leading choice in open AI, and a dip in quality could damage Meta's reputation relative to competitors such as DeepSeek and Anthropic. Furthermore, rumours that management manipulated test results and launched unfinished models raise concerns regarding transparency and reliability within the AI sector.
Who is affected?
The criticism and launch primarily affect AI researchers, open-source developers, and companies that rely on Llama models as the foundation for their own systems and applications.
Impact on the EU
The models were released for global download, but the source mentions no specific regulatory hurdles or adjustments for the EU. Not relevant to EU status.
What else you should know
The allegations that test data had been manipulated originate from an unverified post on the forum 1point3acres and have been further disseminated on Reddit. Meta has reportedly not yet commented on the specific accusations regarding distorted test sets.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Vad har kritiken handlat om?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "Meta defends Llama 4 following bugs and mixed reviews"