The independent survey of AI · Data checked 12 Oct 2026
Every AI model, mapped. Every agent tool, inspected.
180 models placed on one map by price and context. 19,992 agent skills and MCP servers read line by line before they get a verdict. Every number carries its source and the date we checked it.
Cite
Cite this page
ModelsAtlas. “ModelsAtlas — every AI model mapped, every agent tool inspected”. Retrieved 30 Oct 2026. https://modelsatlas.com/
Read the map
Columns A–H are input-price bands, free to $30+ per 1M tokens. Rows 1–6 are context bands, under 32K to over 1M. Every model gets a grid reference; models in the same square are like-for-like.
$0.30–1 input, 1M context. 15 models, among them DeepSeek V4.1 Flash, Gemini 2.5 Flash, Gemini 3 Flash Preview.
$0.10–0.30 input, 256K–1M context. 13 models, among them Ministral 3 14B 2512, Ministral 3 8B 2512, Mistral Small 4.
$0.10–0.30 input, 1M context. 12 models, among them Claude Haiku 5.5, DeepSeek V4 Flash Vision Exp, Gemini 3.1 Flash Lite.
Compare like for like
All 669 comparisons →Claude Opus 4 vs o1
Both cost $15.00 per 1M input tokens. Context: 200K vs 200K.
Claude Opus 4.6 vs Qwen3.8 Max Prime
Qwen3.8 Max Prime costs $4.00 per 1M input tokens against $5.00 for Claude Opus 4.6. Context: 1M vs 1M.
Claude Haiku 4.5 vs Command A
Claude Haiku 4.5 costs $1.00 per 1M input tokens against $2.50 for Command A. Context: 200K vs 256K.
Command R+ (08-2024) vs GPT-4o
Both cost $2.50 per 1M input tokens. Context: 128K vs 128K.
Nano Banana 2.1 vs GPT Audio
Nano Banana 2.1 costs $1.50 per 1M input tokens against $2.50 for GPT Audio. Context: 66K vs 128K.
OpenAI vs Qwen (Alibaba)
Line-ups and prices for 35 and 35 priced models side by side.
Looking for something cheaper?
Lowest prices right now
Cheapest models →| Paid · lowest input | Grid | Input | Output | Context |
|---|---|---|---|---|
| gpt-oss-20bOpenAI | GRIDB3 | $0.02 | $0.09 | 131K |
| Llama 3.2 1B InstructMeta | GRIDB2 | $0.03 | $0.20 | 60K |
| Mistral NemoMistral AI | GRIDB3 | $0.03 | $0.03 | 131K |
| Qwen3.7 FlashQwen (Alibaba) | GRIDB5 | $0.03 | $0.13 | 1M |
| Free tier | Grid | Input | Output | Context |
|---|---|---|---|---|
| Nemotron 3 Ultra (free)NVIDIA | GRIDD5 | Free | Free | 1M |
| Nemotron 3.5 Lightning (free)NVIDIA | GRIDB5 | Free | Free | 1M |
| Gemma 4 26B A4BGoogle | GRIDB4 | Free | Free | 262K |
| Gemma 4 31BGoogle | GRIDB4 | Free | Free | 262K |
Is it safe to install?
How the inspection works →We read every agent skill and MCP server in full before it is listed: filesystem, network, shell, secrets and dependencies. The stamp names the commit we read, and blocked entries stay public, with the reasons.
The three stamps are the most-starred skill or server with each verdict.
Mention Claude Code 2,413 · Cursor 1,353 · OpenClaw 1,114
1,409 servers expose at least one destructive tool. Read-only servers →
Recently audited
Full change log →- Auditverification-before-completion
obra/superpowers· skill ·a56496f✓SAFE B - Auditwriting-plans
obra/superpowers· skill ·a56496f✓SAFE B - Auditdiagnosing-bugs
mattpocock/skills· skill ·743adea✓SAFE B - Auditcontinuous-learning
affaan-m/everything-claude-code· skill ·534c6d9✓SAFE B - Auditcontinuous-learning-v2
affaan-m/everything-claude-code· skill ·534c6d9✓SAFE B - Auditarchitecture-diagram
nousresearch/hermes-agent· skill ·805a6ec✓SAFE B - Auditarxiv
nousresearch/hermes-agent· skill ·805a6ec✓SAFE B - Auditqa
garrytan/gstack· skill ·9693406✓SAFE B
Browse the atlas
19,992 audited listings from 5,302 publishers. Prices for every maker are on the pricing hub.
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- Which AI model with at least 1M tokens of context costs the least per 1M input tokens right now?
- How do I check whether an MCP server is safe to install, and what does the audit look at?
- What is the cheapest alternative to GPT-4 that keeps most of its capabilities?