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 · 1.3M tool calls a day

#	Model	Broken tool calls	Rank	Best on the model	Tool calls a day
1	DeepSeek V4 Pro 0813fp8	0.21%	2 of 21	Ionstream 0.04%	4.4k
2	Kimi K3fp4	0.44%	21 of 24	DeepInfra (mxfp4) 0.01%	28k
3	DeepSeek V4.1 Flash 1017fp8	0.50%	15 of 35	Sail Research (fp4) 0.03%	261k
4	Kimi K2.6int4	0.51%	5 of 14	SiliconFlow (fp8) 0.06%	18k
5	DeepSeek V4.1 Flashfp8	0.66%	19 of 35	Prime Intellect 0.06%	184k
6	Qwen3 235B A22B Instruct 2507fp8	0.78%	1 of 3	—	77k
7	GLM 5.3fp8	0.79%	22 of 42	Amazon Bedrock 0.08%	28k
8	GLM 5.2fp4	0.98%	13 of 27	Friendli 0.15%	9.1k
9	MiniMax M3fp8	1.00%	4 of 11	ModelRun [by Modular] (fp8) 0.27%	9.3k
10	Qwen3.8 27Bfp8	1.02%	14 of 21	Darkbloom (fp4) 0.07%	30k
11	MiniMax M3.1 Flashfp8	1.37%	2 of 11	Together 1.17%	56k
12	GLM 5.3 Flashfp4	1.58%	22 of 34	NextBit (fp4) 0.00%	65k
13	Qwen3.6 35B A3Bfp8	1.75%	2 of 6	AkashML (fp8) 1.63%	96k
14	DeepSeek V4 Flash 0731fp8	1.84%	13 of 22	Sail Research (fp4) 0.04%	135k
15	Qwen3.5 397B A17Bfp8	2.21%	6 of 7	Alibaba Cloud Int. 0.58%	9.1k
16	Gemma 4 31Bfp8	2.23%	8 of 12	ModelRun [by Modular] (fp4) 0.09%	23k
17	DeepSeek V4 Flash 0423fp8	2.32%	10 of 16	Crusoe (fp8) 0.46%	140k
18	gpt-oss-120bfp4	3.47%	7 of 15	Groq 0.04%	21k
19	gpt-oss-20bfp4	22.66%	7 of 7	Darkbloom (fp8) 0.13%	57k

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 model22.66%

of gpt-oss-20b's tool calls fail on Parasail (fp4).

Best model0.21%

on DeepSeek V4 Pro 0813, its best model.

Ignored on2 of 19

models, where the settings tell OpenRouter to skip Parasail.

Fix this with dottxt↵