Price leaderboard
The cheapest frontier LLMs
Ranked by output price per million tokens. For high-volume workloads — classification, extraction, batch inference — the right frontier model can be 10× cheaper than the SOTA at similar quality.
Category leader
#01DeepSeek
DeepSeek: DeepSeek V4 Pro.
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
Output price
$0.87
$ / 1M tokens
Ranking
N=20| # | Model | Provider | Output price | Ctx | Released |
|---|---|---|---|---|---|
| 01 | DeepSeek: DeepSeek V4 Pro | DeepSeek | $0.87 | 1.0M | 2026-04-24 |
| 02 | MiniMax: MiniMax M3 | MiniMax | $1.20 | 1.0M | 2026-05-31 |
| 03 | xAI: Grok 4.3 | xAI | $2.50 | 1.0M | 2026-04-30 |
| 04 | Z.ai: GLM 5.2 | Z.ai | $2.98 | 1.0M | 2026-06-16 |
| 05 | Z.ai: GLM 5.1 | Z.ai | $3.04 | 203K | 2026-04-07 |
| 06 | MoonshotAI: Kimi K2.6 | Moonshot | $3.42 | 262K | 2026-04-20 |
| 07 | NVIDIA: Nemotron 3 Ultra | NVIDIA | $3.60 | 1.0M | 2026-06-04 |
| 08 | MoonshotAI: Kimi K2.7 Code | Moonshot | $3.75 | 262K | 2026-06-12 |
| 09 | Meta: Muse Spark 1.1 | Meta | $4.25 | 1.0M | 2026-07-16 |
| 10 | Qwen: Qwen3.7 Max | Alibaba | $4.42 | 1.0M | 2026-05-21 |
| 11 | xAI: Grok 4.5 | xAI | $6.00 | 500K | 2026-07-08 |
| 12 | Mistral: Mistral Medium 3.5 | Mistral | $7.50 | 262K | 2026-04-30 |
| 13 | MoonshotAI: Kimi K3 | Moonshot | $15.00 | 1.0M | 2026-07-16 |
| 14 | OpenAI: GPT-5.6 Terra | OpenAI | $15.00 | 1.1M | 2026-07-09 |
| 15 | OpenAI: GPT-5.6 Terra Pro | OpenAI | $15.00 | 1.1M | 2026-07-09 |
| 16 | Anthropic: Claude Opus 4.8 | Anthropic | $25.00 | 1.0M | 2026-05-27 |
| 17 | OpenAI: GPT-5.6 Sol | OpenAI | $30.00 | 1.1M | 2026-07-09 |
| 18 | Sakana: Fugu Ultra | Sakana | $30.00 | 1.0M | 2026-06-24 |
| 19 | Anthropic: Claude Fable 5 | Anthropic | $50.00 | 1.0M | 2026-06-09 |
| 20 | Anthropic: Claude Opus 4.8 (Fast) | Anthropic | $50.00 | 1.0M | 2026-05-27 |
Frequently asked
- Which is the cheapest frontier LLM?
- The top of this table shows the model with the lowest output price per 1M tokens, among tracked frontier models. Cheaper models are ideal for high-volume classification, extraction, and batch workloads.
- Do cheaper models cost less overall?
- Not always — reasoning-heavy models generate more tokens per prompt (chain-of-thought, tool calls). Compare cost per completed task, not just per-token price.