Updated October 29, 2026

Tool-call index

Not all models are served equally by all providers. On a typical model, the worst provider breaks tool calls 19× as often as the best. Find the best one for your agents.

openrouter-data-staging.dottxt.workers.dev/llms.txt

← All providers

19 models · 608k tool calls a day

#	Model	Broken tool calls	Rank	Best on the model	Tool calls a day
1	Nemotron 3 Ultrafp8	0.19%	2 of 4	DeepInfra (fp4/flex) 0.17%	11k
2	Qwen3.8 2.4T A95B	0.22%	2 of 5	SiliconFlow (fp8) 0.19%	4.1k
3	Qwen3.8 27Bfp8	0.26%	2 of 21	Darkbloom (fp4) 0.07%	43k
4	MiniMax M3fp8	0.48%	2 of 11	ModelRun [by Modular] (fp8) 0.27%	140k
5	Kimi K2.7 Codeint4	0.53%	5 of 11	NovitaAI (int4) 0.03%	2.8k
6	DeepSeek V4.1 Flashfp8	0.89%	24 of 35	Prime Intellect 0.06%	203k
7	GLM 5.2fp8	1.07%	16 of 27	Friendli 0.15%	1.3k
8	GLM 5.3	1.17%	27 of 42	Amazon Bedrock 0.08%	2.1k
9	Qwen3.5 397B A17B	1.27%	4 of 7	Alibaba Cloud Int. 0.58%	1.2k
10	Kimi K2.6int4	1.41%	10 of 14	SiliconFlow (fp8) 0.06%	2.6k
11	MiMo-V2.6-Flashfp8	1.41%	1 of 9	—	32k
12	GLM 5.3 Flash	1.57%	21 of 34	NextBit (fp4) 0.00%	56k
13	DeepSeek V4 Flash 0731	1.79%	11 of 22	Sail Research (fp4) 0.04%	41k
14	Qwen3 235B A22B Instruct 2507fp8	2.21%	3 of 3	Parasail (fp8) 0.78%	6.8k
15	Kimi K2.5	2.44%	3 of 5	SiliconFlow (int4) 0.17%	8.0k
16	Gemma 4 31Bfp4	2.63%	10 of 12	ModelRun [by Modular] (fp4) 0.09%	2.2k
17	Nemotron 3.5 Super VLfp4	3.44%	4 of 8	Together 0.66%	1.5k
18	DeepSeek V4 Flash 0423	4.33%	15 of 16	Crusoe (fp8) 0.46%	38k
19	gpt-oss-120bint4	17.23%	12 of 15	Groq 0.04%	12k

New models

Get updates when a new model drops

And which provider to use.

For inference providers

Tool calling doesn't have to break

The same provider can handle tool calls well on one model and struggle on the next.

Worst model17.23%

of gpt-oss-120b's tool calls fail on Venice (int4).

Best model0.19%

on Nemotron 3 Ultra, its best model.

Ignored on3 of 19

models, where the settings tell OpenRouter to skip Venice.

Fix this with dottxt↵