What's new
now jobs Jobs market snapshot refreshed now data Recomputed benchmark-weighted quality scores now data Updated speed measurements now data Synced Chatbot Arena benchmark track now data Validated official pricing snapshots now data Pulled latest OpenRouter price index 25 May digest OpenGuardrails: An Open-Source Context-Aware AI Guardrails Platform 25 May digest Published the 2026-05-25 daily digest 25 May digest Workbench Launches Open Source BullMQ Dashboard For Node Backends 24 May digest SpecBench Tests Reward Hacking In Long Horizon Coding Agents now jobs Jobs market snapshot refreshed now data Recomputed benchmark-weighted quality scores now data Updated speed measurements now data Synced Chatbot Arena benchmark track now data Validated official pricing snapshots now data Pulled latest OpenRouter price index 25 May digest OpenGuardrails: An Open-Source Context-Aware AI Guardrails Platform 25 May digest Published the 2026-05-25 daily digest 25 May digest Workbench Launches Open Source BullMQ Dashboard For Node Backends 24 May digest SpecBench Tests Reward Hacking In Long Horizon Coding Agents

AI Speed Comparison

Comprehensive speed analysis — time to first token (TTFT), output speed in tokens per second, and per-provider endpoint performance for models available on multiple platforms.

8

Ultra Fast (100+ tok/s)

12

Fast (50-99 tok/s)

25

Models with TTFT data

1

Multi-provider models

Time to First Token (TTFT)

How quickly the model starts responding. Lower is better — critical for real-time applications and chat interfaces.

# Model TTFT
1 GPT-4.1 Nano OpenAI 150ms
2 Gemini 2.5 Flash Google 160ms
3 Mistral Small 3.1 24B Mistral 180ms
4 GPT-4.1 Mini OpenAI 190ms
5 GPT-4o-mini (2024-07-18) OpenAI 210ms
6 Llama 4 Scout Meta 220ms
7 Claude Sonnet 4.6 Anthropic 240ms
8 GPT-4.1 OpenAI 250ms
9 Llama 4 Maverick Meta 250ms
10 Claude Sonnet 4 Anthropic 260ms
11 Gemini 2.5 Pro Google 270ms
12 Claude Opus 4.6 Anthropic 280ms
13 Llama 3.3 70B Instruct Meta 280ms
14 GPT-4o (2024-05-13) OpenAI 290ms
15 Mistral Large Mistral 310ms
16 GPT-5.2 OpenAI 320ms
17 Qwen2.5 72B Instruct Alibaba 340ms
18 Claude Opus 4 Anthropic 350ms
19 GPT-5 OpenAI 380ms
20 DeepSeek V3.2 DeepSeek 400ms
21 Qwen3 235B A22B Alibaba 420ms
22 DeepSeek R1 DeepSeek 500ms
23 O4 Mini OpenAI 550ms
24 o3 Mini OpenAI 600ms
25 O3 OpenAI 1200ms

Output Speed Rankings

Tokens generated per second. Higher is better for long-form content generation. See full rankings →

1 Gemini 2.5 Flash
350 tok/s
2 GPT-4.1 Nano
200 tok/s
3 GPT-4.1 Mini
160 tok/s
4 GPT-4o-mini (2024-07-18)
150 tok/s
5 Mistral Small 3.1 24B
150 tok/s
6 Llama 4 Scout
120 tok/s
7 GPT-4.1
110 tok/s
8 GPT-4o (2024-05-13)
100 tok/s
9 Llama 4 Maverick
95 tok/s
10 Claude Sonnet 4.6
90 tok/s
11 Gemini 2.5 Pro
90 tok/s
12 GPT-5.2
85 tok/s
13 Claude Sonnet 4
80 tok/s
14 Mistral Large
80 tok/s
15 Llama 3.3 70B Instruct
80 tok/s

Per-Provider Endpoint Comparison

Some open-source models are available on multiple platforms with different speed and pricing. Compare endpoints.

Llama 4 Maverick 2 providers
Provider Speed TTFT
Meta Direct
110 tok/s 180ms
AWS Bedrock
85 tok/s 250ms

About Speed Measurements

TTFT (Time to First Token) — How long until the model starts generating its response, measured in milliseconds. Critical for chat applications where perceived responsiveness matters.

Output Speed (tok/s) — How many tokens the model generates per second once it starts responding. Important for long-form content generation.

Speed data sourced from Artificial Analysis and provider benchmarks. Actual speeds may vary based on load, region, and input complexity.

Fastest AI Models → Full Leaderboard → Compare Models →