Tool Call TTFT (RightRank Probe)

Speed

Time to the first token of a request that must call one tool: median of 5 streamed runs through OpenRouter, pinned to the model's first-party provider, from the us-east-1 region.

Lower is better

Rank	Model	Provider	Time
1	Cohere: Command A+	Cohere	136 ms p90 190 ms · 5 runs
2	Cohere: North Mini Code (free)	Cohere	173 ms p90 204 ms · 5 runs
3	Google: Gemini 2.5 Flash LiteRetiring	Google	306 ms p90 345 ms · 5 runs
4	Inception: Mercury 2	Inception	326 ms p90 355 ms · 5 runs
5	Amazon: Nova Lite 1.0	Amazon	346 ms p90 486 ms · 5 runs
6	Mistral: Ministral 3 3B 2512	Mistral AI	354 ms p90 599 ms · 5 runs
7	Mistral: Mistral Medium 3	Mistral AI	371 ms p90 775 ms · 5 runs
8	OpenAI: GPT-4o-mini	OpenAI	376 ms p90 470 ms · 5 runs
9	Google: Gemini 3.5 Flash Lite	Google	395 ms p90 437 ms · 5 runs
10	Mistral: Mistral Medium 3.1	Mistral AI	403 ms p90 651 ms · 5 runs
11	Mistral: Ministral 3 14B 2512	Mistral AI	427 ms p90 786 ms · 5 runs
12	SpaceXAI: Grok 4.20	SpaceXAI	428 ms p90 651 ms · 5 runs
13	Mistral: Ministral 3 8B 2512	Mistral AI	429 ms p90 468 ms · 5 runs
14	Qwen: Qwen3 30B A3B Thinking 2507Historical	Alibaba	430 ms p90 488 ms · 5 runs
15	Amazon: Nova Micro 1.0	Amazon	431 ms p90 543 ms · 5 runs
16	SpaceXAI: Grok Build 0.1	SpaceXAI	431 ms p90 526 ms · 5 runs
17	Google: Gemini 2.5 FlashRetiring	Google	432 ms p90 529 ms · 5 runs
18	Google: Gemini 3.1 Flash Lite Preview	Google	433 ms p90 583 ms · 5 runs
19	Qwen: Qwen3 14B	Alibaba	435 ms p90 557 ms · 5 runs
20	Mistral: Codestral 2508	Mistral AI	436 ms p90 501 ms · 5 runs
21	Qwen: Qwen3.6 Flash	Alibaba	441 ms p90 499 ms · 5 runs
22	Qwen: Qwen3 30B A3B Instruct 2507	Alibaba	449 ms p90 479 ms · 5 runs
23	Google: Gemini 3.1 Flash Lite	Google	449 ms p90 538 ms · 5 runs
24	Qwen: Qwen3 30B A3BHistorical	Alibaba	467 ms p90 470 ms · 5 runs
25	Qwen: Qwen3 VL 30B A3B ThinkingHistorical	Alibaba	506 ms p90 758 ms · 5 runs
26	Amazon: Nova 2 Lite	Amazon	510 ms p90 736 ms · 5 runs
27	Qwen: Qwen3 Next 80B A3B InstructHistorical	Alibaba	510 ms p90 530 ms · 5 runs
28	Qwen: Qwen3.5-35B-A3B	Alibaba	513 ms p90 996 ms · 5 runs
29	Mistral: Voxtral Small 24B 2507	Mistral AI	513 ms p90 814 ms · 5 runs
30	Cohere: Command R (08-2024)	Cohere	519 ms p90 47.5 s · 5 runs
31	Mistral: Devstral 2 2512	Mistral AI	529 ms p90 543 ms · 5 runs
32	OpenAI: GPT-4o-mini (2024-07-18)	OpenAI	530 ms p90 566 ms · 5 runs
33	Qwen: Qwen3 235B A22B Thinking 2507Historical	Alibaba	532 ms p90 663 ms · 5 runs
34	Qwen: Qwen3.5-122B-A10B	Alibaba	533 ms p90 557 ms · 5 runs
35	MiniMax: MiniMax M3	MiniMax	534 ms p90 690 ms · 5 runs
36	Qwen: Qwen3 VL 8B InstructHistorical	Alibaba	538 ms p90 699 ms · 5 runs
37	OpenAI: GPT-5.4 Mini	OpenAI	548 ms p90 784 ms · 5 runs
38	Qwen: Qwen3.5-27B	Alibaba	550 ms p90 565 ms · 5 runs
39	Qwen: Qwen3 VL 8B ThinkingHistorical	Alibaba	556 ms p90 756 ms · 5 runs
40	OpenAI: GPT-6 Luna	OpenAI	561 ms p90 840 ms · 5 runs
41	Qwen: Qwen3 235B A22BHistorical	Alibaba	563 ms p90 776 ms · 5 runs
42	Mistral: Mistral Large 3 2512	Mistral AI	563 ms p90 729 ms · 5 runs
43	Qwen: Qwen3.8 27B	Alibaba	569 ms p90 866 ms · 5 runs
44	OpenAI: GPT-5.4 Nano	OpenAI	572 ms p90 818 ms · 5 runs
45	Qwen: Qwen3 VL 30B A3B Instruct	Alibaba	573 ms p90 598 ms · 5 runs
46	OpenAI: GPT-3.5 Turbo	OpenAI	575 ms p90 968 ms · 5 runs
47	Qwen: Qwen3 Next 80B A3B ThinkingHistorical	Alibaba	592 ms p90 697 ms · 5 runs
48	OpenAI: GPT-5.6 Luna	OpenAI	602 ms p90 640 ms · 5 runs
49	OpenAI: GPT-4.1 Mini	OpenAI	612 ms p90 642 ms · 5 runs
50	Qwen: Qwen3 235B A22B Instruct 2507	Alibaba	641 ms p90 744 ms · 5 runs

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "Cohere: Command A+",
    "provider": "Cohere",
    "score": "136.0"
  },
  {
    "rank": 2,
    "model": "Cohere: North Mini Code (free)",
    "provider": "Cohere",
    "score": "172.6"
  },
  {
    "rank": 3,
    "model": "Google: Gemini 2.5 Flash Lite",
    "provider": "Google",
    "score": "306.1"
  }
]

Curl

$ curl -s https://rightrank.ai/api/v1/rankings/tool-call-ttft-rightrank-probe.json | jq '.[:3]'

CLI brew install rightrank

$ rightrank top tool-call-ttft-rightrank-probe --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category tool-call-ttft-rightrank-probe
$ rightrank value tool-call-ttft-rightrank-probe  # best performance per dollar

Claude Code Skill Skill

# In Claude Code, use the RightRank skill:
/rightrank best model for tool-call-ttft-rightrank-probe
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on tool-call-ttft-rightrank-probe