Models
Anthropic reportedly releases Claude Fable 5.2 and Mythos 5.2 at lower API prices, with Fable 5.2 first on the Artificial Analysis index
Reports say Claude Fable 5.2 and Mythos 5.2 launched on November 9 at about 20% lower API prices; Fable 5.2 tops one index.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, according to secondary reports that we could read; we did not read Anthropic’s own announcement. The reported API price is $8 per million input tokens and $40 per million output tokens, against $10 and $50 for version 5.1, a cut of about 20%. On the Artificial Analysis Intelligence Index, Fable 5.2 scored 65 and ranked first among 74 models in a snapshot taken November 10. A claim that it leads Qwen4-Max by 22 points is not supported by anything we found.
The essentials
- Release: reported November 9, 2026, with API names claude-fable-5-2 and claude-mythos-5-2. Fable 5.2 needs no access approval; Mythos 5.2 is limited to organisations verified through Anthropic’s life-sciences and cyber programmes.
- Price: $8 input and $40 output per million tokens, against $10 and $50 for 5.1. A fast mode costs $16 and $80. The 20% figure is the calculation of one report.
- Index: Fable 5.2 (Max setting) scored 65 on Artificial Analysis Intelligence Index v5.0.1. Gemini 4 Carbon is reported about 5 points behind.
- Cost of the lead: the same page reports about $26.54 and 46 minutes per index task at Max, so the headline score comes at a high running cost.
The 22-point claim
The request said Fable 5.2 leads Qwen4-Max by 22 points. We found no Qwen4-Max score. The reports we read name Gemini 4 Carbon as the nearest rival, about 5 points behind, and compare against earlier Qwen models. The 22-point figure may come from a different comparison or a different snapshot, so treat it as unconfirmed.
What ‘lower price’ means here
The cut applies to list prices per token. Because Fable 5.2 is reported to be slow and verbose, with many output tokens per task, the cost of finishing a task can still be high. List price and cost per task are different measures.
What is not established
- Anthropic’s own announcement: not read by us.
- A Qwen4-Max score and the 22-point gap: not found.
- Whether the index lead holds on other benchmarks: one report says GPT-6 Astra is ahead on a science index.
- Whether the lower price is permanent or promotional: not stated.
- Mythos 5.2 benchmark results: not reported.
Summary table
| Point | What is stated as of November 7, 2026 | Source |
|---|---|---|
| Release date | November 9, 2026 (reported) | Braindetox |
| Input price | $8 per million tokens (5.1: $10) | Braindetox |
| Output price | $40 per million tokens (5.1: $50) | Braindetox |
| Index score | 65, first of 74 models (Max setting) | Artificial Analysis |
| Nearest rival | Gemini 4 Carbon, about 5 points behind | OfficeChai |
FAQ
Is the 22-point lead over Qwen4-Max confirmed?
No. We found no Qwen4-Max score. Reports put the lead over the next model, Gemini 4 Carbon, at about 5 points.
How much cheaper are the new models?
Reports give $8 and $40 per million input and output tokens, against $10 and $50 before, about 20% less.
Can anyone use Mythos 5.2?
Reportedly not. It is limited to organisations verified through Anthropic’s programmes. Fable 5.2 has no approval step.
Method and limits
Checked on November 16, 2026. We read Braindetox, Artificial Analysis, OfficeChai and BuildMVPFast through page fetches and search excerpts, not Anthropic’s announcement. The request called the sources secondary and unverified with Anthropic; that is right. The 22-point gap with Qwen4-Max is not supported by what we found.
Sources
- Braindetox: Claude Fable 5.2 and Mythos 5.2 release (verified 2026-11-08)
- Artificial Analysis: Claude Fable 5.2 (verified 2026-11-08)
- OfficeChai: Anthropic jumps to top of Artificial Analysis Intelligence Index with Fable 5.2 (verified 2026-11-08)
- BuildMVPFast: Claude Fable 5.2 and Mythos 5.2: what changed for builders (verified 2026-11-08)
Spotted an error in this article? Report it · Our corrections policy