Which models cap at 128K?

Smaller frontier models (GPT-4o-mini at some tiers, Claude Haiku on legacy), mid-tier open source (Llama 3.3, some Mistral variants), older Qwen and DeepSeek lines. Many newer models ship 128K as baseline with 200K+ extended.

Is the cheapest 128K model good enough for RAG?

Usually yes. RAG typically sends 4K to 32K retrieved context. A 128K window is 4x to 30x what you actually need, so the lowest-priced 128K model is often the optimal RAG backbone.

Does 128K mean accurate up to 128K?

No. Many 128K models lose fidelity past 64K. Gemini and Claude hold up best. If retrieval quality matters, use RAG with short context and pick a reasoning-strong model.

Why pay more at this tier?

Premium 128K models (Opus, GPT-5) win on quality per token, tool use, and output speed. If you care about latency or reasoning quality, pay up. If you just need a big window, pick the cheap one.

Context · 128K+

Cheapest 128K context LLMs

Every LLM with a 128,000+ token context window. Ranked by input price per 1M tokens.

Models40

Cheapest$0.00

Min context128K tokens

All pricing Pricing home

What this page is

128K is the modern baseline for LLM context. Every model priced per 1M tokens at 128K+ is listed here, cheapest first. For RAG pipelines, long conversations, and mid-sized documents, this is the sweet spot.

Ranked by input price

128K+ context models, cheapest first.

#	Model	Provider	In $/1M	Out $/1M	Context	context	Type
1	ERNIE-5.5-Flash-Preview (free)	baidu	$0.00	$0.00	1.0M	1.0M	Closed
2	Qwen3 Next 80B A3B Instruct (free)	Alibaba Qwen	$0.00	$0.00	262K	262K	OSS
3	Nemotron 3.5 Lightning (free)	NVIDIA	$0.00	$0.00	1.0M	1.0M	Closed
4	P Laguna XS.2 (free)	poolside	$0.00	$0.00	262K	262K	OSS
5	gpt-oss-120b (free)	OpenAI	$0.00	$0.00	131K	131K	OSS
6	Llama Guard 4 12B (free)	Meta	$0.00	$0.00	164K	164K	Closed
7	Trinity Large Preview (free)	arcee-ai	$0.00	$0.00	131K	131K	OSS
8	MiniMax M2.7 (free)	minimax	$0.00	$0.00	197K	197K	Closed
9	DeepSeek V4 Flash 0731 (free)	DeepSeek	$0.00	$0.00	1.0M	1.0M	Closed
10	Gemma 4 31B (free)	Google DeepMind	$0.00	$0.00	262K	262K	OSS
11	MiniMax M3 (free)	minimax	$0.00	$0.00	1.0M	1.0M	Closed
12	Nex-N2-Pro (free)	nex-agi	$0.00	$0.00	262K	262K	OSS
13	Nex-N2.5-Pro (free)	nex-agi	$0.00	$0.00	262K	262K	Closed
14	Nemotron 3.5 Super VL (free)	NVIDIA	$0.00	$0.00	1.0M	1.0M	Closed
15	LFM3-1.7B (free)	liquid	$0.00	$0.00	262K	262K	Closed
16	gpt-oss-20b (free)	OpenAI	$0.00	$0.00	131K	131K	OSS
17	Nemotron 3 Nano 30B A3B (free)	NVIDIA	$0.00	$0.00	256K	256K	OSS
18	Qwen3.8 27B (free)	Alibaba Qwen	$0.00	$0.00	262K	262K	Closed
19	Hy3 preview (free)	tencent	$0.00	$0.00	262K	262K	Closed
20	Step 3.5 Flash (free)	stepfun	$0.00	$0.00	256K	256K	OSS
21	Nemotron Nano 9B V2 (free)	NVIDIA	$0.00	$0.00	128K	128K	OSS
22	Lyria 3 Pro Preview	Google DeepMind	$0.00	$0.00	1.0M	1.0M	Closed
23	MiniMax M3.1 Flash (free)	minimax	$0.00	$0.00	1.0M	1.0M	Closed
24	Nemotron 3 Super (free)	NVIDIA	$0.00	$0.00	262K	262K	OSS
25	GLM 4.5 Air (free)	z-ai	$0.00	$0.00	131K	131K	OSS
26	Llama 3.2 3B Instruct (free)	Meta	$0.00	$0.00	131K	131K	OSS
27	Mistral Small 3.1 24B (free)	Mistral AI	$0.00	$0.00	128K	128K	OSS
28	Lyria 3 Clip Preview	Google DeepMind	$0.00	$0.00	1.0M	1.0M	Closed
29	Gemma 4 26B A4B (free)	Google DeepMind	$0.00	$0.00	262K	262K	OSS
30	MiniMax M2.5 (free)	minimax	$0.00	$0.00	197K	197K	OSS
31	P Laguna M.1 (free)	poolside	$0.00	$0.00	262K	262K	OSS
32	North Mini Code (free)	Cohere	$0.00	$0.00	256K	256K	OSS
33	R Flint	Recursive	$0.00	$0.00	262K	262K	OSS
34	Owl Alpha	openrouter	$0.00	$0.00	1.0M	1.0M	Closed
35	Nemotron 3 Nano Omni (free)	NVIDIA	$0.00	$0.00	256K	256K	OSS
36	Gemma 3 27B (free)	Google DeepMind	$0.00	$0.00	131K	131K	OSS
37	Nex-N2.5-Mini (free)	nex-agi	$0.00	$0.00	262K	262K	Closed
38	Llama 3.3 70B Instruct (free)	Meta	$0.00	$0.00	131K	131K	OSS
39	Elephant	openrouter	$0.00	$0.00	262K	262K	Closed
40	Trinity Mini (free)	arcee-ai	$0.00	$0.00	131K	131K	OSS

Top 3 cheapest 128K context LLMs

Cheapest 128K

ERNIE-5.5-Flash-Preview (free)

ERNIE-5.5-Flash-Preview (free) offers 1.0M context at $0.00/M input. Baseline modern model.

Runner up · 128K

Qwen3 Next 80B A3B Instruct (free)

Qwen3 Next 80B A3B Instruct (free) offers 262K context at $0.00/M input. Baseline modern model.

Third · 128K

Nemotron 3.5 Lightning (free)

Nemotron 3.5 Lightning (free) offers 1.0M context at $0.00/M input. Baseline modern model.

The price gap · cheapest vs most expensive

Cheapest

ERNIE-5.5-Flash-Preview (free)

$0.00/M

$ per 1M input tokens

Why the gap

At 128K, premium pricing pays for reasoning quality and vendor reliability, not window size. For RAG backbones, the cheap end almost always wins.

Most expensive

ERNIE-5.5-Flash-Preview (free)

$0.00/M

$ per 1M input tokens

Frequently asked questions

For most production use cases, yes. 128K fits 300+ pages, a full API spec, or a large function library. Only reach for 200K+ when you regularly exceed 100K token prompts.

Cheapest 128K context LLMs

Ranked by input price

Top 3 cheapest 128K context LLMs

The price gap · cheapest vs most expensive

Frequently asked questions

See also

Other context tiers

Stacks

Compare