Matched a much pricier model on a LeetCode-style set, but degraded noticeably once the prompt exceeded ~60k tokens.
source: news.ycombinator.com ↗AUG 23, 2026
DeepSeek · released APR 24, 2026
Open-weights reasoning model at an aggressive price point.
Matched a much pricier model on a LeetCode-style set, but degraded noticeably once the prompt exceeded ~60k tokens.