Get API key

Rankings

Every live model on llm.buzz, by what people use and what we measure: tokens, speed, latency, price and uptime. View all models

Usage data: no complete week published yet

Speed, latency and uptime through Oct 27, 04:00 UTC

Live models
81
Speed measured
50
95%+ uptime
49

Fastest models

Median output tokens per second over the last 24 hours, from each model's fastest live provider. Top 10 of 50 models with a figure.

  1. 1Anthropic: Claude Sonnet 5.5on Anthropic · compare997 tok/s test100% uptime
  2. 2OpenAI: gpt-oss-120bon llm.buzz · compare481 tok/s test100% uptime
  3. 3OpenAI: GPT-6 Lunaon llm.buzz · compare401 tok/s test100% uptime
  4. 4Anthropic: Claude Opus 5.5on Anthropic · compare334 tok/s test100% uptime
  5. 4Moondream: Moondream 3 Previewon Prime Intellect334 tok/s test98.6% uptime
  6. 4Moondream: Moondream 3.1on Prime Intellect334 tok/s test100% uptime
  7. 4OpenAI: GPT-4.1on llm.buzz · compare334 tok/s test100% uptime
  8. 8Anthropic: Claude Fable 5on Anthropic · compare278 tok/s test100% uptime
  9. 8Anthropic: Claude Fable 5.1on Anthropic · compare278 tok/s test100% uptime
  10. 8Z.ai: GLM 5.3on Prime Intellect · compare278 tok/s test100% uptime

Quickest first token

Median time to first token over the last 24 hours, from each model's quickest live provider. Top 10 of 50 models with a figure.

  1. 1Surogate: Surogate 3.7 Flashby Surogate · compare0.04 s test100% uptime
  2. 2Mistral: Mistral Small 3.2 24Bby Mistral · compare0.19 s test100% uptime
  3. 3Google: Gemma 4 26B A4Bby Google · compare0.23 s test100% uptime
  4. 3OpenAI: gpt-oss-120bby OpenAI · compare0.23 s test100% uptime
  5. 3Qwen: Qwen3.8 27Bby Qwen · compare0.23 s test100% uptime
  6. 6Moondream: Moondream 3.1by Moondream0.28 s test100% uptime
  7. 6Qwen: Qwen3.6 35B A3Bby Qwen · compare0.28 s test100% uptime
  8. 6Z.ai: GLM 5.3by Z.ai · compare0.28 s test100% uptime
  9. 9Moondream: Moondream 3 Previewby Moondream0.33 s test98.6% uptime
  10. 10DeepSeek: DeepSeek V4 Flash 0731by DeepSeek · compare0.40 s test100% uptime

Lowest price

The first-ranked provider's input and output price per million tokens, ranked by their sum. Top 10 of 50 models with a figure.

  1. 1OpenAI: gpt-oss-120bby OpenAI · compare$0.05 / $0.25100% uptime
  2. 2OpenAI: GPT-5 Nanoby OpenAI · compare10% off $0.045 / $0.36100% uptime
  3. 3Google: Gemma 4 26B A4Bby Google · compare$0.07 / $0.34100% uptime
  4. 4Xiaomi: MiMo-V2.6-Flashby Xiaomi$0.14 / $0.2897.3% uptime
  5. 5Mistral: Mistral Small 3.2 24Bby Mistral · compare25% off $0.129675 / $0.30255100% uptime
  6. 6Google: Gemma 4 31Bby Google · compare25% off $0.15 / $0.30100% uptime
  7. 7Qwen: Qwen3.8 Flashby Qwen · compare25% off $0.1125 / $0.3525100% uptime
  8. 8NVIDIA: Nemotron 3 Superby NVIDIA$0.085 / $0.4052.1% uptime
  9. 9Anthropic: Claude Haiku 5.5by Anthropic · compare$0.10 / $0.50100% uptime
  10. 9DeepSeek: DeepSeek V4.1 Flash 1017by DeepSeek · compare25% off $0.15 / $0.45100% uptime

Most providers

Providers serving the model right now, each checked against every capability the model declares. Top 10 of 81 models with a figure.

Benchmarks
  1. 1MoonshotAI: Kimi K3by MoonshotAI · compare6100% uptime
  2. 2Z.ai: GLM 5.3by Z.ai · compare5100% uptime
  3. 3Qwen: Qwen3.8 27Bby Qwen · compare4100% uptime
  4. 3Qwen: Qwen3.8 Flashby Qwen · compare4100% uptime
  5. 3Z.ai: GLM 5.3 Flashby Z.ai · compare4100% uptime
  6. 6DeepSeek: DeepSeek V4 Flash 0731by DeepSeek · compare3100% uptime
  7. 6DeepSeek: DeepSeek V4.1 Flashby DeepSeek · compare3100% uptime
  8. 6DeepSeek: DeepSeek V4.1 Flash 1017by DeepSeek · compare3100% uptime
  9. 6Google: Gemma 4 26B A4Bby Google · compare3100% uptime
  10. 6OpenAI: gpt-oss-120bby OpenAI · compare3100% uptime

Speed against price

Each model's best median output speed over 24 hours against its first-ranked provider's blended price (3 input to 1 output). The dashed line joins the models no other beats on both.

04008001200$0.10$0.30$1.00$3.00$10.00$30.00Blended $ per 1M tokens, log scaleAnthropic: Claude Fable 5: $15.00 blended, 278 tok/s (test requests)Anthropic: Claude Fable 5.1: $15.00 blended, 278 tok/s (test requests)Anthropic: Claude Haiku 5.5: $0.20 blended, 193 tok/s (test requests)Anthropic: Claude Opus 4.7: $7.50 blended, 65 tok/s (test requests)Anthropic: Claude Opus 4.8: $7.50 blended, 65 tok/s (test requests)Anthropic: Claude Opus 5: $7.50 blended, 232 tok/s (test requests)Anthropic: Claude Opus 5.5: $6.00 blended, 334 tok/s (test requests)Anthropic: Claude Sonnet 5: $3.00 blended, 78 tok/s (test requests)Anthropic: Claude Sonnet 5.5: $4.00 blended, 997 tok/s (test requests)DeepSeek: DeepSeek V4 Flash 0731: $0.5762 blended, 134 tok/s (test requests)DeepSeek: DeepSeek V4 Flash Vision Exp: $0.495 blended, 193 tok/s (test requests)DeepSeek: DeepSeek V4 Pro 0813: $1.63 blended, 65 tok/s (test requests)DeepSeek: DeepSeek V4.1 Flash: $0.525 blended, 193 tok/s (test requests)DeepSeek: DeepSeek V4.1 Flash 1017: $0.225 blended, 193 tok/s (test requests)Google: Gemma 4 26B A4B: $0.1375 blended, 232 tok/s (test requests)Google: Gemma 4 31B: $0.1875 blended, 65 tok/s (test requests)Mistral: Mistral Medium 3.5: $3.46 blended, 65 tok/s (test requests)Mistral: Mistral Small 3.2 24B: $0.1729 blended, 134 tok/s (test requests)Moondream: Moondream 3 Preview: $0.85 blended, 334 tok/s (test requests)Moondream: Moondream 3.1: $0.85 blended, 334 tok/s (test requests)MoonshotAI: Kimi K3: $5.13 blended, 78 tok/s (test requests)MoonshotAI: Kimi K3.1 Pro: $4.00 blended, 37 tok/s (test requests)NVIDIA: Nemotron 3 Super: $0.1638 blended, 193 tok/s (test requests)OpenAI: GPT-4.1: $3.50 blended, 334 tok/s (test requests)OpenAI: GPT-4.1 Mini: $0.63 blended, 134 tok/s (test requests)OpenAI: GPT-4o-mini: $0.2625 blended, 93 tok/s (test requests)OpenAI: GPT-5 Mini: $0.6875 blended, 93 tok/s (test requests)OpenAI: GPT-5 Nano: $0.1237 blended, 161 tok/s (test requests)OpenAI: GPT-5.4: $5.63 blended, 65 tok/s (test requests)OpenAI: GPT-5.4 Mini: $1.69 blended, 134 tok/s (test requests)OpenAI: GPT-5.4 Nano: $0.4625 blended, 193 tok/s (test requests)OpenAI: GPT-5.5: $11.25 blended, 232 tok/s (test requests)OpenAI: GPT-5.6 Luna: $0.45 blended, 232 tok/s (test requests)OpenAI: GPT-5.6 Sol: $8.00 blended, 78 tok/s (test requests)OpenAI: GPT-5.6 Terra: $4.05 blended, 93 tok/s (test requests)OpenAI: GPT-6 Astra: $20.00 blended, 31 tok/s (test requests)OpenAI: GPT-6 Luna: $0.20 blended, 401 tok/s (test requests)OpenAI: GPT-6 Sol: $4.00 blended, 93 tok/s (test requests)OpenAI: GPT-6.1 Sol: $4.00 blended, 31 tok/s (test requests)OpenAI: gpt-oss-120b: $0.10 blended, 481 tok/s (test requests)Qwen: Qwen3.6 35B A3B: $0.50 blended, 161 tok/s (test requests)Qwen: Qwen3.8 27B: $0.775 blended, 161 tok/s (test requests)Qwen: Qwen3.8 Flash: $0.1725 blended, 78 tok/s (test requests)Qwen: Qwen3.8 Max (0902): $3.00 blended, 45 tok/s (test requests)Qwen: Qwen3.8 Omni Flash: $0.23 blended, 78 tok/s (test requests)Surogate: Surogate 3.7 Flash: $0.2625 blended, 112 tok/s (test requests)Xiaomi: MiMo-V2.6-Flash: $0.175 blended, 13 tok/s (test requests)Xiaomi: MiMo-V2.6-Pro: $0.54 blended, 37 tok/s (test requests)Z.ai: GLM 5.3: $1.50 blended, 278 tok/s (test requests)Z.ai: GLM 5.3 Flash: $0.2375 blended, 93 tok/s (test requests)
  1. 1OpenAI: gpt-oss-120bby OpenAI481 tok/s test$0.10 blended
  2. 2Anthropic: Claude Sonnet 5.5by Anthropic997 tok/s test$4.00 blended

Newest models

The latest models to go live on llm.buzz, with the day each went public.

  1. 1Google: Nano Banana Pro (Gemini 3 Pro Image)by GoogleOct 25, 2026
  2. 2BAAI: bge-base-en-v1.5by BAAIOct 25, 2026
  3. 3BAAI: bge-large-en-v1.5by BAAIOct 25, 2026
  4. 4BAAI: bge-m3by BAAIOct 25, 2026
  5. 5Black Forest Labs: FLUX.2 Klein 4Bby Black Forest LabsOct 25, 2026

Uptime

Live models by how often they answered over the last 72 hours, failover included.

Status
  1. 1NVIDIA: Nemotron 3 Superby NVIDIA52.1%

Usage rankings

Usage is published weekly, once at least 5 organizations use a model. The current week closes Monday, Nov 2, 2026, 00:00 UTC.

What appears here once a week publishes

  1. Top modelsInput plus output tokens per complete week in UTC: the five most used models, then everyone else.
  2. LeaderboardModels by published tokens, with the change against the period before.
  3. Market shareEach maker's share of published tokens: the nine largest, then everyone else.
  4. Open and closed weightsLast week's published tokens on models with open weights against closed ones. A model under the publishing floor adds nothing.
  5. By capabilityLast week's published tokens on models that can call tools, reason, read images or take 200K tokens of context. A model counts in each it has; these are the models' capabilities, not what each request used.
  6. Top appsApps that call llm.buzz, by published tokens.

Browse all 67 apps and agents that work with llm.buzz

How these are measured

What each figure counts, over which window, and when a figure is withheld.

What is measured?

Output speed and time to first token are medians over the last 24 hours, from each model's best live provider. Uptime is how often the model answered over the last 72 hours, failover included. Every figure comes from llm.buzz's test requests: short streamed requests and periodic full ones sent to each live provider.

Which price?

The price of the provider a request goes to first, per million tokens. The price board ranks by input plus output.

Where is usage?

Usage rankings by tokens appear here once the first complete week of published usage closes. Until then this page ranks only what llm.buzz measures.

Is this a quality ranking?

No. These figures measure speed, availability, price and use, not answer quality.

What counts?

Input plus output tokens of every request through the Chat Completions, Responses and Messages APIs, as the console counts them.

Which weeks?

Complete ISO weeks in UTC. The current week appears once it ends. For a model public for less than eight weeks, the 4-week figures count only the weeks since it became public.

Who shows up?

A model or app appears only when at least 5 organizations used it that week and no single one sent more than 80% of its tokens.

Call any model on the listOne key for every model on llm.buzz, each request served by a live provider.
Get API key