llama.cpp
The C++ inference engine powering most local LLMs.
OPEN SOURCECPU-CAPABLE
DeepSeek
DeepSeek V4 Flash
Absurdly cheap open-weights workhorse — $0.10/$0.20 per 1M.
Open weightstextREL 2026-04-24
Compare this model →Verified 2026-06-12
Overview
DeepSeek V4 Flash is the budget king: $0.10/$0.20 per 1M tokens for a capable 1M-context open-weights model. It is the default model for this site's own automation pipeline — good enough for structured drafting at near-zero cost.
Benchmark scores
Methodology →Intelligence47
Codingn/a
Reasoningn/a
Mathn/a
Sources and verification
- Intelligence: Artificial Analysis Intelligence Index · checked 2026-06-12
- Catalogue and pricing: OpenRouter model record · checked 2026-06-12
Strengths and weaknesses
Strengths
- Cheapest capable frontier-family model
- Open weights
- 1M context
Trade-offs
- Clear quality gap to V4 Pro
- Text-only
Reach for this model when
High-volume drafting, extraction and automation where cost rules.
Run something like it locally
Self-host quantised builds via llama.cpp / Ollama on 24GB+ GPUs.