Best Value AI Models
DeepSeek V4.1 Flash tops the Value LLMs chart with $0.28 per 1M in. GPT-4o-mini is second with $0.15 per 1M in.
Frontier models are not always the smartest buy. This chart ranks the ten best-value large language models by cost-performance — capability delivered per dollar of inference. A small open-weights model that is 90% as good for 10% of the price outranks a frontier model on value. Prices shown are indicative input rates per 1M tokens and shift often; treat them as a snapshot.
DeepSeek V4.1 Flash
Input $0.28 / output $0.42 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
GPT-4o-mini
Input $0.15 / output $0.60 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Gemini 2.5 Flash
Input $0.30 / output $2.50 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Qwen3-235B
Input $0.20 / output $0.40 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Llama 4 Maverick
Input $0.20 / output $0.40 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Grok 2 mini
Input $0.20 / output $0.40 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
GLM-4.5
Input $0.20 / output $0.40 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Claude Haiku 4.5
Input $1.00 / output $5.00 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Mistral Large
Input $0.50 / output $1.50 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Phi-4
Input $0.10 / output $0.15 per 1M tokens (indicative). Ranked by cost-performance, not raw benchmark — cheap models that punch above their price lead.
Editor’s Notes
- Ranking is editorial by cost-performance (capability per dollar), not a single benchmark score.
- Prices are input rates per 1M tokens, refreshed daily by fetch-ai.js from the litellm price library; falls back to a hardcoded snapshot on failure.
- Prices change frequently — verify before quoting.
Last updated: 2026-11-02