Anthropic released Claude Fable 5.2 on November 9, and Mythos 5.2 became available at the same time, but only to verified organizations. Fable scored 49.1% on the FrontierCode 2.0 coding benchmark, higher than Opus 5.5's 45.5%; its input price dropped from $10 to $8 per million tokens, and output is about 30% faster. Interestingly, Anthropic said that it deliberately kept the capabilities most likely to accelerate AI research at the level of Opus 5.5 instead of making them available. The stronger the models get, the more obvious these holdbacks become. Do you think this is responsible, or are they afraid things will move too fast?
#热点背后的利益 #Claude