All General Coding Reasoning Multimodal Safety Specialized Agentic Language Long Context Speed Text-to-Image Image Editing Text-to-Video Image-to-Video
Generation Tokens/sec incl. Reasoning (RightRank Probe)113 Inter-Token Latency Latency Output Tokens/sec163 Output Tokens/sec (RightRank Probe)165 TTFT163 TTFT (RightRank Probe)180 Time to First Answer Token (RightRank Probe)100 Tool Call TTFT (RightRank Probe)169 Tool Call Total Time (RightRank Probe)169 Total Time (RightRank Probe)181

Output Tokens/sec (RightRank Probe)

Speed

Answer tokens per second after the first token, for a prompt that asks for about 500 tokens: median of 5 streamed runs through OpenRouter, pinned to the model's first-party provider, from the us-east-1 region.

Higher is better

Rank Model Provider Throughput
1 Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) Google
547.3 tok/s p90 628.2 tok/s · 5 runs
2 OpenAI: GPT-5 Image Mini OpenAI
477.0 tok/s p90 497.7 tok/s · 5 runs
3 Google: Gemini 3.7 Flash Google
447.7 tok/s p90 523.0 tok/s · 5 runs
5 SpaceXAI: Grok 4.20 Multi-Agent SpaceXAI
396.3 tok/s p90 1195.2 tok/s · 5 runs
6 Inception: Mercury 2 Inception
387.1 tok/s p90 434.3 tok/s · 4 runs
7 Google: Gemini 3.8 Flash Google
364.3 tok/s p90 471.1 tok/s · 5 runs
8 OpenAI: GPT-5 Mini OpenAI
330.1 tok/s p90 476.6 tok/s · 5 runs
9 Google: Gemini 3.5 Flash Google
319.6 tok/s p90 347.9 tok/s · 5 runs
10 Amazon: Nova Micro 1.0 Amazon
305.3 tok/s p90 323.6 tok/s · 5 runs
11 Mistral: Ministral 3 3B 2512 Mistral AI
301.3 tok/s p90 308.0 tok/s · 5 runs
12 OpenAI: o3 Mini OpenAI
300.8 tok/s p90 501.3 tok/s · 5 runs
18 ↳OpenAI: o3 Mini High OpenAI
251.1 tok/s p90 251.1 tok/s · 1 runs
13 OpenAI: o4 Mini OpenAI
287.4 tok/s p90 485.3 tok/s · 5 runs
4 ↳OpenAI: o4 Mini High OpenAI
441.7 tok/s p90 451.2 tok/s · 3 runs
14 Google: Gemini 2.5 Flash LiteRetiring Google
284.0 tok/s p90 299.6 tok/s · 4 runs
15 Z.ai: GLM 5.2 Z.ai
273.3 tok/s p90 292.1 tok/s · 5 runs
16 OpenAI: GPT-6 Luna OpenAI
256.8 tok/s p90 309.6 tok/s · 5 runs
17 Z.ai: GLM 5 Z.ai
256.0 tok/s p90 324.1 tok/s · 5 runs
19 Google: Gemini 3.1 Flash Lite Google
246.6 tok/s p90 267.5 tok/s · 5 runs
20 Google: Gemini 3.1 Flash Lite Preview Google
245.5 tok/s p90 262.3 tok/s · 5 runs
21 Google: Gemini 3.5 Flash Lite Google
242.2 tok/s p90 248.2 tok/s · 5 runs
22 Google: Gemini 3.6 Flash Google
229.5 tok/s p90 254.8 tok/s · 5 runs
23 Google: Nano Banana 2 (Gemini 3.1 Flash Image) Google
212.4 tok/s p90 215.3 tok/s · 5 runs
24 Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview) Google
210.2 tok/s p90 230.5 tok/s · 5 runs
25 Amazon: Nova 2 Lite Amazon
209.9 tok/s p90 228.2 tok/s · 5 runs
26 Perplexity: Sonar Perplexity
205.9 tok/s p90 230.9 tok/s · 5 runs
27 Google: Gemini 2.5 FlashRetiring Google
197.1 tok/s p90 208.8 tok/s · 5 runs
28 OpenAI: GPT-5.4 Nano OpenAI
195.6 tok/s p90 204.9 tok/s · 5 runs
29 OpenAI: GPT-4.1 Nano OpenAI
194.6 tok/s p90 201.4 tok/s · 5 runs
30 Qwen: Qwen3.5-Flash Alibaba
194.5 tok/s p90 208.3 tok/s · 5 runs
31 Amazon: Nova Lite 1.0 Amazon
193.1 tok/s p90 222.9 tok/s · 5 runs
32 Qwen: Qwen3 Coder Next Alibaba
183.6 tok/s p90 191.8 tok/s · 5 runs
33 OpenAI: GPT-4.1 Mini OpenAI
178.4 tok/s p90 183.1 tok/s · 5 runs
34 OpenAI: o1 OpenAI
175.0 tok/s p90 204.7 tok/s · 5 runs
35 Google: Gemini 3 Flash Preview Google
172.7 tok/s p90 185.3 tok/s · 5 runs
36 Amazon: Nova Pro 1.0 Amazon
171.3 tok/s p90 184.5 tok/s · 5 runs
37 Mistral: Voxtral Small 24B 2507 Mistral AI
165.7 tok/s p90 174.3 tok/s · 5 runs
38 OpenAI: GPT-4o (2024-05-13) OpenAI
165.6 tok/s p90 170.1 tok/s · 5 runs
39 Qwen: Qwen3.5-122B-A10B Alibaba
164.9 tok/s p90 173.4 tok/s · 5 runs
40 Anthropic: Claude Opus 5 Anthropic
163.6 tok/s p90 316.0 tok/s · 5 runs
41 SpaceXAI: Grok Build 0.1 SpaceXAI
160.3 tok/s p90 191.3 tok/s · 5 runs
42 Google: Gemini 3.1 Pro Preview Google
158.5 tok/s p90 172.4 tok/s · 5 runs
43 OpenAI: GPT-4o OpenAI
158.3 tok/s p90 183.0 tok/s · 5 runs
44 OpenAI: GPT-3.5 Turbo 16k OpenAI
154.4 tok/s p90 160.3 tok/s · 5 runs
45 OpenAI: GPT-4o-mini OpenAI
153.9 tok/s p90 164.3 tok/s · 5 runs
46 Qwen: Qwen3 VL 8B InstructHistorical Alibaba
152.6 tok/s p90 155.9 tok/s · 5 runs
47 Mistral: Mistral Medium 3.1 Mistral AI
151.0 tok/s p90 154.6 tok/s · 5 runs
48 Google: Gemini 3.1 Pro Preview Custom Tools Google
150.8 tok/s p90 172.0 tok/s · 5 runs
49 Mistral: Mistral Medium 3 Mistral AI
148.7 tok/s p90 154.6 tok/s · 5 runs
50 Qwen: Qwen3 VL 8B ThinkingHistorical Alibaba
148.5 tok/s p90 149.1 tok/s · 5 runs

Access this data

JSON GET /api/v1 
[
  {
    "rank": 1,
    "model": "Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)",
    "provider": "Google",
    "score": "547.3"
  },
  {
    "rank": 2,
    "model": "OpenAI: GPT-5 Image Mini",
    "provider": "OpenAI",
    "score": "477.0"
  },
  {
    "rank": 3,
    "model": "Google: Gemini 3.7 Flash",
    "provider": "Google",
    "score": "447.7"
  }
]
Curl 
$ curl -s https://rightrank.ai/api/v1/rankings/output-tokens-sec-rightrank-probe.json | jq '.[:3]'
CLI brew install rightrank 
$ rightrank top output-tokens-sec-rightrank-probe --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category output-tokens-sec-rightrank-probe
$ rightrank value output-tokens-sec-rightrank-probe  # best performance per dollar
Claude Code Skill Skill 
# In Claude Code, use the RightRank skill:
/rightrank best model for output-tokens-sec-rightrank-probe
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on output-tokens-sec-rightrank-probe