Unverified50% confidenceTradeoffExact time
Both cache read and cache write token counts are included in determining Haiku 5.5's 100K pricing threshold, so caching cannot help avoid the pricing cliff.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
UnverifiedThe minimum cacheable length for Haiku 5.5 is 512 tokens, and prompts shorter than this will not error but will not be cached; a cache entry only becomes available after the first response begins; by default the cache lives for 5 minutes and is refreshed for free on each hit.74% similarUnverifiedClaude Haiku 5.5's context window has expanded from Haiku 4.5's 200,000 tokens to 1 million tokens, and the output limit has been raised from 64K to 128K.70% similarUnverifiedEmpirically, above 20 token/s is close to reading speed and allows smooth interaction, while below 5 token/s is noticeably laggy.67% similarUnverifiedOpus 5 输出 Token 单价约为 Haiku 的数十倍，因此压缩输出 Token 对高端模型的降本效果远比低端模型显著64% similarUnverifiedThe autoregressive generation phase is a memory-bandwidth-sensitive workload: generating each token requires reading the model weights in full once (or at least reading the weights involved in the activation path), so the actual speed is approximately equal to the VRAM bandwidth divided by the weight volume that needs to be read per token.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1054198API
curl https://kongchang.com/api/v1/knowledge/claims/1054198MCP
get_claim(id=1054198)