Browse AI models

18 models tracked - grouped by family, best families first

Families Newest Quality Value Size
All Edge Consumer Enthusiast Workstation Premier
Any 8GB 12GB 24GB 32GB 48GB 64GB 96GB
More filters

Jeff

mstrasser (firelex) 1 model best 82
Jeff Sep 28, 2026
2.0B params 262k ctx 1GB min RAM apache 2.0 2 var

Nemotron

🇺🇸 NVIDIA 2 models best 80
Nemotron 3 Diarization Sep 23, 2026
100M params 0GB min RAM openmdw 1.1 2 var
Nemotron 3 Nano 4B Jun 5, 2026
4.0B 1000k ctx 3GB min RAM other 1 var

Gliner

🇺🇸 Fastino 1 model best 64
GLiNER2.5-Decide Sep 24, 2026
340M params 1GB min RAM apache 2.0 2 var

Gemma

🇺🇸 Google 5 models best 58
Gemma 3 270M Aug 14, 2025 no marks
270M 32k ctx 1GB min RAM other 2 var
25
Gemma 3n E2B Jun 26, 2025 no marks
2.0B 32k ctx 6GB min RAM other 2 var
48
Gemma 3n E4B Jun 26, 2025 no marks
4.0B 32k ctx 8GB min RAM other 2 var
58
Gemma 3 1B Feb 27, 2025 no marks
1.0B 32k ctx 1GB min RAM other 2 var
38
Gemma 3 4B Feb 27, 2025 no marks
4.3B 128k ctx 3GB min RAM other 2 var $0.05/M in
55
1100 pts/$

Qwen

🇨🇳 Alibaba 1 model best 56
Qwen3 4B Apr 28, 2025 text + metadata
4.0B 32k ctx 3GB min RAM apache 2.0 1 var
56

Phi

🇺🇸 Microsoft 1 model best 52
Phi-4 Mini Feb 1, 2025 no marks
3.8B 128k ctx 3GB min RAM mit 2 var
52

Llama

🇺🇸 Meta 2 models best 50
Llama 3.2 1B Instruct Sep 25, 2024 no marks
1.2B 128k ctx 1GB min RAM llama 3 2 var $0.03/M in
35
1296 pts/$
Llama 3.2 3B Instruct Sep 25, 2024 no marks
3.2B 128k ctx 2GB min RAM llama 3 2 var $0.05/M in
50
1000 pts/$

Cua S1

trycua 1 model best 0
CUA-S1-FORMS Sep 18, 2026
no local build mit (source only no weights published)

Jev

TypeSafe 1 model best 0
Jev Sep 15, 2026 no marks
60k ctx no local build proprietary $0.04/M in

Laya

ConvAI Innovations 1 model best 0
Laya Sep 18, 2026 no marks
420M params 1k ctx 1GB min RAM apache 2.0 3 var

Needle

Cactus Compute 1 model best 0
Needle 3 Sep 17, 2026
120M params 8k ctx no local build

Openjev

Undisclosed 1 model best 0
OpenJev Sep 16, 2026 no marks
600M params 8k ctx 1GB min RAM mit 3 var

FAQ

What is the cheapest LLM API?

Open-weight models like DeepSeek, Qwen, and GLM are the cheapest LLM APIs - often under $1 per million input tokens. Sort the table by best value to rank quality per dollar.

Which LLM has the best coding score?

Coding scores vary by benchmark; filter the table by use case coding to see coding_score per model, or open any model's page for its benchmark breakdown.

How do I compare LLM pricing?

This table shows per-million-token input and output prices across providers. Open a model's pricing page for live rates, cached-input discounts, and price history.