Use case · Writing

Cheapest writing LLMs

Best-value models for general writing · blogs, marketing, drafts, summaries. Ranked by input token price with quality score.

Models30
Cheapest$0.01
ScopeLLM · Multimodal
What this page is
This page ranks general-purpose LLMs (text and multimodal chat models) for writing workloads. For bulk writing, value per dollar is what matters: the quality gap between the cheapest and the most expensive model is much smaller than the price gap. Our "top picks" use a quality-to-price ratio, not just raw benchmarks. For specialized coding or reasoning writing, see the dedicated pages.

General-purpose LLMs, cheapest first.

#ModelIn $/1MOut $/1MType
1liquid logoLFM2-2.6B$0.01$0.02OSS
2liquid logoLFM2-8B-A1B$0.01$0.02OSS
3Ling-2.6-flash$0.01$0.03Closed
4DeepSeek logoDeepSeek V4 Flash 0731$0.02$1.28Closed
5ibm-granite logoGranite 4.0 Micro$0.02$0.11OSS
6OpenAI logogpt-oss-20b$0.02$0.09OSS
7Mistral AI logoMistral Nemo$0.02$0.03OSS
8Meta logoLlama 3.1 8B Instruct$0.02$0.03OSS
9OpenAI logogpt-oss-20b (batch)$0.02$0.11Closed
10DeepSeek logoDeepSeek V4 Flash$0.03$1.28OSS
11OpenAI logoGPT-5 Nano (batch)$0.03$0.20Closed
12nex-agi logoNex-N2-Mini$0.03$0.10Closed
13nex-agi logoNex-N2.5-Mini$0.03$0.10Closed
14Meta logoLlama 3.2 1B Instruct$0.03$0.20OSS
15Google DeepMind logoGemma 2 9B$0.03$0.09OSS
16liquid logoLFM2-24B-A2B$0.03$0.12OSS
17Alibaba Qwen logoQwen2.5 Coder 7B Instruct$0.03$0.09OSS
18Alibaba Qwen logoQwen3.7 Flash$0.03$0.13Closed
19Alibaba Qwen logoQwen-Turbo$0.03$0.13OSS
20Amazon logoNova Micro 1.0$0.04$0.14Closed
21OpenAI logogpt-oss-120b$0.04$0.17OSS
22Cohere logoCommand R7B (12-2024)$0.04$0.15Closed
23sao10k logoLlama 3 8B Lunaris$0.04$0.05OSS
24inception logoMercury 2.5$0.04$0.15Closed
25inception logoMercury 2.5 Preview$0.04$0.15Closed
26NVIDIA logoNemotron Nano 9B V2$0.04$0.16OSS
27tencent logoHy-MT2-1.8B$0.04$0.18Closed
28arcee-ai logoTrinity Mini$0.04$0.15OSS
29Alibaba Qwen logoQwen3 30B A3B Instruct 2507$0.05$0.19OSS
30OpenAI logoGPT-6 Luna (batch)$0.05$0.25Closed
Cheapest
LFM2-2.6B
$0.01/M
$ per 1M input tokens
Why the gap

The premium on writing models is almost entirely brand voice and tone. The quality gap on factual accuracy, grammar, and structure is within noise for most real-world writing jobs.

Most expensive
GPT-6 Luna (batch)
$0.05/M
$ per 1M input tokens
Not in 2026. Models like Gemini Flash, Claude Haiku, GPT-4.1 mini, and DeepSeek V3 produce polished prose at a fraction of premium prices. For routine drafts, newsletters, and social posts, there is almost no quality gap.