LLMAtlas

Workspace

  • Overview

Develop

  • Playground
  • Code
    Soon
  • Projects
    Soon
  • Compare
  • Prompts

Research

  • Models
  • Cost Calculator
  • Leaderboard
  • API Status

Learn

  • Learn
    New

Account

  • Settings
Sign in free

Save prompts, track progress, and sync your workspace across devices.

Sign in· Register
LLMAtlas

Workspace

  • Overview

Develop

  • Playground
  • Code
    Soon
  • Projects
    Soon
  • Compare
  • Prompts

Research

  • Models
  • Cost Calculator
  • Leaderboard
  • API Status

Learn

  • Learn
    New

Account

  • Settings
Sign in free

Save prompts, track progress, and sync your workspace across devices.

Sign in· Register
⌘K
Global Leaderboard
Cached· synced 0s ago

The Global LLM Leaderboard

Every model in the LLMAtlas catalogue, ranked by a composite of quality, speed, context, openness and value — synced live with the OpenRouter catalogue and scored against industry benchmarks. Inspired by openrouter.ai/rankings.

Models
0
Free
0
Frontier
0
Providers
0
Avg context
0K
New / 45d
0

Top 5 — capability radar

Composite scoring methodology

Quality (35%)

Composite benchmark — MMLU-Pro, GPQA, HumanEval, SWE-bench, MATH & LMArena. Real where published, otherwise derived.

Speed (20%)

Measured throughput by provider hardware. Groq & Cerebras top the charts with LPU / wafer-scale silicon.

Context (15%)

Log-normalised context window. 2M tokens = 100, 1M ≈ 91, 128K ≈ 64, 8K ≈ 28.

Openness (15%)

100 for fully open weights (MIT/Apache), 30 for closed/proprietary. Open = reproducible, auditable, self-hostable.

Value (15%)

100 for free tiers, scaled inversely with price per token. Free APIs score 100; frontier-paid ~28.

Showing 50 of 201 models · ranked by composite

#ModelUsage Composite Quality Speed Context Access
12
GR

Llama 4 Maverick (Groq)

Meta · Groq

29B/wk
93
90
88
1.0M
FREE
26
GR

Llama 4 Scout (Groq)

Meta · Groq

23B/wk
92
87
95
524K
FREE
32
NV

Llama 4 Maverick 17B (128E)

Meta · NVIDIA NIM

28B/wk
92
90
84
1.0M
FREE
43
GI

Llama 4 Maverick (free via GitHub)

Meta · GitHub Models

18B/wk
91
90
80
1M
FREE
510
GI

Llama 4 Scout (free via GitHub)

Meta · GitHub Models

23B/wk
90
87
84
524K
FREE
66
PO

Llama 4 Maverick (keyless)

Meta · Pollinations

25B/wk
90
90
74
1.0M
FREE
75
GR

Llama 3.3 70B (Versatile)

Meta · Groq

16B/wk
89
86
97
128K
FREE
812
PO

Llama 4 Scout (keyless)

Meta · Pollinations

21B/wk
89
87
82
524K
FREE
95
GR

DeepSeek R1 Distill Llama 70B

DeepSeek · Groq

17B/wk
88
84
95
128K
FREE
1010
TO

Llama 3.3 70B Turbo (free)

Meta · Together AI

19B/wk
88
86
90
131K
FREE
110
OP

Llama 4 Scout

Meta · OpenRouter

15B/wk
88
87
82
10M
PAID
1211
GR

Qwen3 32B (Groq)

Alibaba · Groq

15B/wk
88
82
96
128K
FREE
135
GR

QwQ 32B

Alibaba · Groq

18B/wk
87
81
94
128K
FREE
147
CE

GPT-OSS 120B (Cerebras)

OpenAI · Cerebras

19B/wk
87
87
100
33K
FREE
158
NV

Llama 3.3 70B (NVIDIA)

Meta · NVIDIA NIM

15B/wk
87
86
86
131K
FREE
1612
NV

Llama 3.3 Nemotron Super 49B

Nvidia · NVIDIA NIM

21B/wk
87
87
85
131K
FREE
176
PO

DeepSeek Pro (keyless)

DeepSeek · Pollinations

24B/wk
87
93
70
164K
FREE
186
PO

Qwen Large (keyless)

Alibaba · Pollinations

23B/wk
87
90
72
262K
FREE
194
PO

Qwen Coder (keyless)

Alibaba · Pollinations

16B/wk
87
86
80
262K
FREE
203
DE

DeepSeek V4 Flash

DeepSeek · DeepSeek

17B/wk
87
86
88
131K
FREE
218
OP

Qwen 3.6 Flash

Alibaba · OpenRouter

19B/wk
87
88
82
1M
PAID
221
NV

DeepSeek V4 Pro (NVIDIA)

DeepSeek · NVIDIA NIM

27B/wk
87
94
65
262K
FREE
236
NV

Qwen 3.6 (NVIDIA)

Alibaba · NVIDIA NIM

19B/wk
87
92
68
262K
FREE
249
GR

DeepSeek R1 Distill Qwen 32B (Groq)

DeepSeek · Groq

15B/wk
87
80
97
128K
FREE
255
OP

DeepSeek V4 Flash

DeepSeek · OpenRouter

14B/wk
86
86
82
1.0M
PAID
263
TO

Llama 4 Maverick

Meta · Together AI

18B/wk
86
90
82
524K
PAID
273
NV

Llama 3.1 70B (NVIDIA)

Meta · NVIDIA NIM

13B/wk
86
83
86
131K
FREE
283
NV

DeepSeek V4 Flash (NVIDIA)

DeepSeek · NVIDIA NIM

17B/wk
86
86
82
131K
FREE
294
GI

Llama 3.3 70B (free via GitHub)

Meta · GitHub Models

15B/wk
86
86
80
131K
FREE
301
PO

GPT-OSS 20B Fast (Pollinations, keyless)

OpenAI · Pollinations

22B/wk
86
79
96
128K
FREE
311
PO

DeepSeek (keyless)

DeepSeek · Pollinations

26B/wk
86
88
80
131K
FREE
320
OP

Qwen 3.6 Plus

Alibaba · OpenRouter

21B/wk
86
91
72
1M
PAID
3312
OP

Llama 4 Maverick

Meta · OpenRouter

16B/wk
86
90
74
1.0M
PAID
344
GI

DeepSeek V4 Pro (free via GitHub)

DeepSeek · GitHub Models

20B/wk
86
94
60
262K
FREE
354
CE

Qwen3 32B (Cerebras)

Alibaba · Cerebras

13B/wk
86
82
100
33K
FREE
360
GR

Llama 3.2 90B Vision

Meta · Groq

16B/wk
85
82
85
128K
FREE
372
OP

Qwen3 Next 80B A3B (free)

Alibaba · OpenRouter

14B/wk
85
84
80
131K
FREE
381
CE

GLM-4.7 (Cerebras)

ZAI · Cerebras

9B/wk
85
80
99
33K
FREE
399
NV

Nemotron Nano 9B v2

Nvidia · NVIDIA NIM

16B/wk
85
74
95
131K
FREE
404
NV

Qwen3 Next 80B A3B (NVIDIA)

Alibaba · NVIDIA NIM

21B/wk
85
84
80
131K
FREE
4111
PO

GLM (keyless)

ZAI · Pollinations

6B/wk
85
84
80
131K
FREE
424
OP

Qwen3 Coder Flash

Alibaba · OpenRouter

13B/wk
85
87
78
1M
PAID
437
CE

DeepSeek R1 Distill Qwen 32B (Cerebras)

DeepSeek · Cerebras

16B/wk
85
80
100
33K
FREE
448
GR

Llama 3.1 8B Instant

Meta · Groq

12B/wk
84
70
99
128K
FREE
450
GR

Llama 3.2 11B Vision

Meta · Groq

10B/wk
84
74
92
128K
FREE
461
OP

Llama 3.3 70B Instruct (free)

Meta · OpenRouter

21B/wk
84
82
78
131K
FREE
476
OP

GPT-OSS 20B (free)

OpenAI · OpenRouter

18B/wk
84
79
83
131K
FREE
485
OP

Nemotron 3 Super 120B (free)

Nvidia · OpenRouter

21B/wk
84
89
65
131K
FREE
499
OP

Qwen3 Coder (free)

Alibaba · OpenRouter

17B/wk
84
83
78
131K
FREE
506
OP

GLM-4.5 Air (free)

ZAI · OpenRouter

6B/wk
84
78
84
131K
FREE

Composite scores and capability dimensions are approximate. Benchmark values are reported from vendor cards & public evals where available, otherwise derived from the composite and flagged accordingly. Usage share, tokens/week and weekly trend are modeled estimates — OpenRouter does not expose per-model token volumes publicly. Model catalogue, pricing and context windows for OpenRouter models are synced live; 12 upcoming models are speculative projections.