Anthropic announced the launch of Claude Opus 4.8 on Wednesday, cutting prices for fast mode by two-thirds. The company indicated that its most powerful model, Mythos, will be available to all users within weeks.
The Opus 4.8 benchmark price remains fixed at $5 in and $25 out per million tokens, Per Anthropic Blog.
Opus 4.8 undercuts GPT-5.5 and outperforms it on most benchmarks
Fast mode now costs $10 input and $50 output per million tokens at 2.5x speed. In Opus 4.7, the price of the same level reached $30/$150. At standard rates, Opus 4.8 charges $25 per million output tokens. The price of GPT-5.5 is $30.
| Standard | Opus 4.7 | Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro |
|---|---|---|---|---|
| Sui-Seats Pro | 64.3% | 69.2% | 58.6% | 54.2% |
| SWE seat verified | 87.6% | 88.6% | — | — |
| I use 2026 math | 69.3% | 96.7% | — | — |
| Terminal 2.1 | 66.1% | 74.6% | — | — |
| GraphWalks F1 (1 million tokens) | 40.3% | 68.1% | — | — |
| Online-Mind2Web | — | 84% | less than 84% | — |
GPT-5.5 features a terminal workflow and CLI. VentureBeat reports that Opus 4.8 outperforms GPT-5.5 in at least 12 benchmarks including cognitive work, use of proxy tools, and long-context tasks.
The enterprise’s partners confirmed gains in production. Databricks reported an “incremental change in proxy logic” within its Genie data proxy “at a nominal cost of 61% cheaper than Opus 4.7” through multimedia efficiency on PDFs and graphs.
Cognition said that in 4.8 the comment verbosity and widget invocation issues from 4.7 were fixed. Hippia noted more pronounced quotation accuracy in dense financial depositions.
The degree of imbalance in Opus 4.8 is on par with Mythos
The alignment team at Anthropic ran about 2,600 simulation realizations using Opus 4.8. They found the misalignment rate to be around 1.9 compared to 2.5 for Opus 4.7. This value is approximately equal to the Mythos Preview value.
The model is four times less likely than 4.7 to allow defects in generated code to go unreported. It received a 0% non-critical reporting of defective results, the first Claude model to reach this mark.
The anthropologist noted one concern. Anthropic has identified one area of concern. In about 5% of training cases, the model began thinking about the evaluation criteria without being informed that it was being evaluated.
The company said this did not result in observable worse behavior, but called it a “worrying trend that may complicate training in the future,” according to VentureBeat.
Mythos Preview remains limited to a small number of organizations within the Glasswing cybersecurity business. Anthropic said it expects to offer Mythos-class models to all of its customers in the coming weeks, once additional cyber safeguards are in place, according to Axios. The company has also teased cheaper models that will offer “many of the same capabilities as the Opus.”
like Cryptopolitan noted When Opus 4.5 was released back in November, Anthropic was on a release schedule roughly every two months since 2026. The arrival of Opus 4.8 just 41 days after Opus 4.7 has accelerated that schedule.
The DeepSeek gap widened in the same week
Two days ago, DeepSeek permanently reduced V4-Pro production prices to $0.87 per million tokens. Standard production of the Opus 4.8 is $25. Express mode is $50.
Cryptocurrency trading bots and DeFi agents process millions of tokens every session. Given these quantities, DeepSeek’s cost of $0.87 makes sense. Opus’s $25 fee doesn’t do that.
Don’t just read cryptocurrency news. Understand that. Subscribe to our newsletter. It’s free.
