DeepSeek: DeepSeek v4.1 Flash 1017
Points and comments are a snapshot, not live.
DeepSeek V4.1 Flash 1017 offers low-cost, high-throughput AI with strong coding benchmarks.
DeepSeek V4.1 Flash 1017 is available on OpenRouter at $0.10/M input and $0.50/M output tokens. It supports 1M context and a weighted average price of $0.01264/M input, $0.574/M output. Performance: 150 tok/s throughput, 0.56s latency, 99.91% availability. Benchmarks include #30 on Artificial Analysis Intelligence Index (39.5), #7 on AA-LCR (84.0%), and #20 on Terminal-Bench 4.0 (26.8%). Providers offer varied pricing and routing options. Top apps: Claude Code, Hermes Agent, and pi.
What commenters are saying
Users praise DeepSeek V4.1 Flash for cost efficiency and speed, especially vs US providers. Some note productivity gains from fast throughput and openness (visible CoT, no random blocks). A split exists: heavy users burn through $200 subscriptions in hours with suboptimal agent workflows, while others find it hard to reach caps. Criticism: web search API gets mixed results. Benchmarks show it trails Opus 5.5 and GPT-6 Astra but beats older models. Many favor using DeepSeek for coding with a stronger model (Opus) for planning/review.