DeepSeek-V4.1-Flash-1017 release: open weights, 1M context and pricing
A summary of the official release: DeepSWE 77.0 (from 74.2) and Terminal-Bench 42.3 (from 31.2), MIT open weights, 1M context, API at $0.15/$0.60 per 1M tokens off-peak (2x at peak), and Search/Fetch at $0.75 per 1k calls.
🚨 DEEPSEEK'S FLASH GOT SMARTER AND FASTER, AND STILL COSTS CENTS 🪙⚡ Update: DeepSeek just shipped a new V4.1 Flash snapshot, DeepSeek-V4.1-Flash-1017: • DeepSeek says it beats the previous version on benchmarks, finishes tasks faster and gives shorter answers (a "Pareto improvement") • Live on web, app and API • Open-weights on Hugging Face (MIT license) • 1M context, OpenAI- and Anthropic-compatible API formats • API: $0.15 in / $0.60 out per 1M tokens off-peak (weekends count as off-peak); peak is 2x • Also on OpenRouter and Vercel AI Gateway • Same thread: Search and Fetch APIs are open to everyone, $0.75 per 1k calls each • According to DeepSeek: DeepSWE 77.0 (was 74.2) and Terminal-Bench 42.3 (was 31.2) Caveats: • Benchmarks are DeepSeek's own numbers, not independent tests • Weekday peak hours (01-04 and 06-10 UTC) cost 2x • Search/Fetch calls through Claude Code, Codex and DeepSeek Harness are free for now; billing starts Oct 31 🇧🇷 O Flash da DeepSeek ficou mais esperto e mais rápido, e continua custando centavos. Cheap enough to make you switch?