AITOPAI热报
Models

Claude Haiku 幻觉问题遭吐槽，性能被 GPT-5.6-Luna 超越

Official accountSimon Willison@simonw
Post
View the post
Why we picked it

Simon Willison 说 Claude Haiku 幻觉多，不如 GPT-5.6-Luna，还会污染 Claude Code 抓网页，用起来要小心。

Simon Willison 在推文中称 Claude Haiku 是他当前最不喜欢的模型。他认为该模型幻觉问题严重，性能已被同价位的 GPT-5.

Body

6-Luna 超越。他还指出 Claude Code WebFetch 工具仍在调用 Claude Haiku，导致每次抓取 URL 都可能产生幻觉。

该推文获得 196 次点赞和 9928 次浏览。

Original · Simon Willison

Claude Haiku is my current least favorite model - it hallucinates wildly, and is out-performed now by other similarly priced models like GPT-5.6-Luna Even worse: it seems to still be used by the Claude Code WebFetch tool, which means hallucination risk any time you fetch a URL!📊 43 ⚡

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Claude Haiku, GPT-5.6-Luna, Claude Code, hallucinations.

Source attribution: curated from Simon Willison (@simonw).

#Claude Haiku#GPT-5.6-Luna#Claude Code#hallucinations

Keep reading

Recent stories with the same tag or section
ModelsUS vs ChinaMulti-source08:36
Chinese contentGoogle 发布 Gemini 4 Hydrogen，llm-gemini 0.37 同步更新支持

Google 的便宜版 Gemini 4 来了，但现在的定价只比老大哥便宜 25%，要到 12 月才真正划算。文章里有完整的三模型价格表和 pelican 输出质量对比，选模型前值得看看实际效果差距。

#Gemini 4#Google#Claude Haiku#model pricing
Event
Official accountSimon Willison’s WeblogSource
ModelsUS vs ChinaMulti-source71°16:07
Chinese contentMeta 发布 Muse Ember：主打智能体高频任务的小型推理模型

Meta 这次没做大模型，做了个专给智能体干脏活的小模型，缓存输入只要 $0.004，便宜到几乎白送。用 Artificial Analysis 的图看，同样的任务成本它最省。跑 Agent、做批量文档抽取的团队可以重点看看定价表那一栏。

#Muse Ember#Meta#Claude Haiku#GPT-6 Luna
Event
Official accountSimon Willison’s WeblogSource
ModelsUS vs ChinaMulti-source72°04:44
Chinese contentMeta 发布 Muse Ember：对标 Claude Haiku 5.5，成本约三分之一

Meta 这个 Muse Ember 是专门给 agent 干杂活用的：工具调用、代码子任务、路由抽取这些高频小任务，价格只有 Claude Haiku 的三分之一左右。做 agent 流水线的开发者可以直接算一下迁移成本，省的钱是实打实的。

#Muse Ember#Meta#Claude Haiku#agents
Event
rohanpaul_ai@rohanpaul_aiSource
ModelsUS vs ChinaMulti-sourceSelected71°03:18
Meta Muse Ember (Max) tops Code Arena WebDev small-model ranking at 1608 pts

The benchmark gap over Haiku 5.5 and GPT-6 Luna is modest in points but wide in cost per completed task, which matters most for teams running high-volume codegen. Developers building website or app scaffolding pipelines can test Ember against Haiku on their own workloads to validate the claimed 2x cost advantage at equal API rates. Volume API buyers should watch whether actual task-level cost, not sticker pricing, becomes the new battleground for small coding models.

#Muse Ember#Meta#Claude Haiku#GPT-6 Luna
Event
lmarena.ai@lmarena_aiSource
ModelsUS vs ChinaMulti-sourceSelected02:27
Muse Ember Small Model Hits Efficiency Frontier at One-Third of Haiku Cost

The interesting shift here is that small-model competition is now argued in cost-per-task terms rather than benchmark leaderboards, a framing that maps directly onto procurement decisions. Teams running high-volume workloads like classification, extraction, and batch processing should re-run their cost baselines against Haiku before renewing API commitments. The luna comparison also hints at a tiered small-model lineup that deserves watching as pricing details surface.

#muse ember#Scale AI#Claude Haiku#small language models
Event
Official accountAlexandr Wang (Meta MSL)@alexandr_wangSource