Inference Index
Live · Updated 1d ago

AI intelligence,
indexed.

The definitive directory for the AI ecosystem. Seven directories. Hundreds of entries. Tracked, compared, and explained by people who actually ship with this stuff.

View leaderboard →Cost calculatorModel match →
Models tracked
22
Providers
21
Interfaces
20
Directories
7
The index pulse · ii-v1.0

Inference Index rose 0.00% this week as cost index was flat while intelligence index was flat.

View trends →
Inference Index
100.00
0.00% / 7d
Intelligence
1353.2
0.00% / 7d
Cost
$7.253
0.00% / 7d
Value
186.6
0.00% / 7d
PRICEClaude Opus 4.7 · in $15.00 · out $75.00·
PRICEClaude Sonnet 4.6 · in $3.00 · out $15.00·
PRICEClaude Haiku 4.5 · in $0.80 · out $4.00·
PRICEGPT-5 Turbo · in $6.00 · out $24.00·
PRICEGPT-5 Mini · in $0.50 · out $2.00·
PRICEo4 · in $15.00 · out $60.00·
PRICEGemini 2.5 Ultra · in $7.00 · out $21.00·
PRICEGemini 2.5 Flash · in $0.30 · out $1.20·
PRICELlama 4 Behemoth · in $2.50 · out $8.00·
PRICELlama 4 Scout · in $0.35 · out $0.80·
PRICEMistral X-Large · in $2.00 · out $6.00·
PRICEDeepSeek V4 · in $0.14 · out $0.28·
BENCHClaude Opus 4.7 · LMSYS ELO 1412·
BENCHClaude Sonnet 4.6 · LMSYS ELO 1368·
BENCHClaude Haiku 4.5 · LMSYS ELO 1305·
BENCHGPT-5 Turbo · LMSYS ELO 1398·
BENCHGPT-5 Mini · LMSYS ELO 1288·
BENCHo4 · MMLU-Pro 93.8%·
BENCHGemini 2.5 Ultra · LMSYS ELO 1385·
BENCHGemini 2.5 Flash · LMSYS ELO 1312·
INDEXInference Index tracks 23 models · 22 providers · 20 interfaces · 18 harnesses·
PRICEClaude Opus 4.7 · in $15.00 · out $75.00·
PRICEClaude Sonnet 4.6 · in $3.00 · out $15.00·
PRICEClaude Haiku 4.5 · in $0.80 · out $4.00·
PRICEGPT-5 Turbo · in $6.00 · out $24.00·
PRICEGPT-5 Mini · in $0.50 · out $2.00·
PRICEo4 · in $15.00 · out $60.00·
PRICEGemini 2.5 Ultra · in $7.00 · out $21.00·
PRICEGemini 2.5 Flash · in $0.30 · out $1.20·
PRICELlama 4 Behemoth · in $2.50 · out $8.00·
PRICELlama 4 Scout · in $0.35 · out $0.80·
PRICEMistral X-Large · in $2.00 · out $6.00·
PRICEDeepSeek V4 · in $0.14 · out $0.28·
BENCHClaude Opus 4.7 · LMSYS ELO 1412·
BENCHClaude Sonnet 4.6 · LMSYS ELO 1368·
BENCHClaude Haiku 4.5 · LMSYS ELO 1305·
BENCHGPT-5 Turbo · LMSYS ELO 1398·
BENCHGPT-5 Mini · LMSYS ELO 1288·
BENCHo4 · MMLU-Pro 93.8%·
BENCHGemini 2.5 Ultra · LMSYS ELO 1385·
BENCHGemini 2.5 Flash · LMSYS ELO 1312·
INDEXInference Index tracks 23 models · 22 providers · 20 interfaces · 18 harnesses·
The seven directories

The complete taxonomy of AI.

Every layer of the stack — from the models themselves, to the providers that serve them, to the interfaces and harnesses built on top. All indexed.

01 /22

Models

Every frontier and open-weight model, ranked by ELO, MMLU, and $/token.

Claude Opus 4.7GPT-5 TurboGemini 2.5 UltraClaude Sonnet 4.6→
02 /21

Providers

Labs, aggregators, alt-stacks, and self-host. Where inference actually lives.

AnthropicOpenAIGoogle AI StudioxAI→
03 /20

Interfaces

Chat apps, code editors, CLIs, agents — every way humans meet models.

Claude.aiChatGPTGeminiPerplexity→
04 /18

Harnesses

Orchestration frameworks, agent runtimes, and multi-agent platforms.

CrewAILangGraphAutoGenSemantic Kernel→
05 /15

Benchmarks

The evals themselves, explained — what they measure and why they matter.

MMLU-ProHumanEvalSWE-BenchLMSYS Chatbot Arena→
06 /20

Plugins & Tools

MCP servers, GPT actions, function-calling tools — the agent toolkit.

GitHub MCP ServerFilesystem MCPSupabase MCP ServerSlack MCP→
07 /18

Datasets

Pretraining, instruction, and eval data — the fuel behind every model.

FineWebThe Stack v2Common CrawlRedPajama v2→
Top ranked · by LMSYS ELO

The leaderboard.

Full leaderboard →
Frontier cost curve

The price of a thought

Blended list price for flagship models (input + 3×output, per 1M tokens). Log scale — every gridline is a 10× move.

GPT flagship
Claude flagship
Gemini flagship
2022-112023-052023-112024-042024-102025-042025-092026-03$13$33$90$240
Tools · cost calculator

How much will it cost?

Drag, compare, decide. Real pricing across every model we track. Shareable, exportable.

Scenario
10,000
600
400
Total tokens / day
10.00M tokens
Model/ day/ month
Full calculator + share →
Intelligence · signal, not noise

This week in AI.

All intelligence →
ReleaseNaNy ago

Claude Opus 4.7 ships — 1M context, 72% SWE-Bench, new world record

Anthropic’s latest flagship arrives with a 1M-token context window and the highest SWE-Bench score ever recorded on a public model.

Read →
AnalysisNaNy ago

DeepSeek V4 quietly reset the cost curve again

ReleaseNaNy ago

Google ships Gemini 2.5 Flash with 1M context at $0.30/M

IndustryNaNy ago

The MCP ecosystem is growing faster than the GPT Store ever did