Model Compatibility

AgentNode Runtime tested against 182 models from 32 providers. 175 pass all 4 scenarios (96%).

Last tested: 2026-04-07

182
Models Tested
175
Perfect Score (S-Tier)
32
Providers
96%
S-Tier Rate

Tier System

S
Perfect
4/4 scenarios passed
175
A
Great
3/4 scenarios passed
2
B
Partial
2/4 scenarios passed
1
C
Minimal
1/4 scenarios passed
3
F
Unsupported
0/4 scenarios passed
1

Test Scenarios

1Capabilities

Model calls agentnode_capabilities to list installed tools

2Search + Install

Model searches for a package, then installs it

3Run Tool

Model executes a tool with specific arguments

4Multi-step

Model chains multiple tool calls autonomously

182 of 182 models
ProviderModelTierS1S2S3S4Score
alibabatongyi-deepresearch-30b-a3bS4/4
allenaiolmo-3.1-32b-instructS4/4
amazonnova-2-lite-v1S4/4
amazonnova-lite-v1S4/4
amazonnova-micro-v1S4/4
amazonnova-premier-v1S4/4
amazonnova-pro-v1S4/4
anthropicclaude-3-haikuS4/4
anthropicclaude-3.7-sonnetS4/4
anthropicclaude-haiku-4.5S4/4
anthropicclaude-opus-4S4/4
anthropicclaude-opus-4.1S4/4
anthropicclaude-opus-4.5S4/4
anthropicclaude-opus-4.6S4/4
anthropicclaude-sonnet-4S4/4
anthropicclaude-sonnet-4.5S4/4
anthropicclaude-sonnet-4.6S4/4
bytedance-seedseed-1.6S4/4
bytedance-seedseed-1.6-flashS4/4
bytedance-seedseed-2.0-liteS4/4
bytedance-seedseed-2.0-miniS4/4
coherecommand-r-08-2024S4/4
coherecommand-r-plus-08-2024S4/4
deepseekdeepseek-chatS4/4
deepseekdeepseek-chat-v3-0324S4/4
deepseekdeepseek-chat-v3.1S4/4
deepseekdeepseek-r1S4/4
deepseekdeepseek-v3.1-terminusS4/4
deepseekdeepseek-v3.2S4/4
deepseekdeepseek-v3.2-expS4/4
essentialairnj-1-instructS4/4
googlegemini-2.0-flash-001S4/4
googlegemini-2.0-flash-lite-001S4/4
googlegemini-2.5-flashS4/4
googlegemini-2.5-flash-liteS4/4
googlegemini-2.5-proS4/4
googlegemma-4-26b-a4b-itS4/4
googlegemma-4-31b-itS4/4
inceptionmercuryS4/4
inceptionmercury-2S4/4
inceptionmercury-coderS4/4
kwaipilotkat-coder-pro-v2S4/4
meta-llamallama-3-8b-instructS4/4
meta-llamallama-3.1-70b-instructS4/4
meta-llamallama-3.1-8b-instructS4/4
meta-llamallama-3.3-70b-instructS4/4
meta-llamallama-4-maverickS4/4
meta-llamallama-4-scoutS4/4
minimaxminimax-m1S4/4
minimaxminimax-m2S4/4
minimaxminimax-m2.1S4/4
minimaxminimax-m2.5S4/4
minimaxminimax-m2.7S4/4
mistralaicodestral-2508S4/4
mistralaidevstral-2512S4/4
mistralaidevstral-mediumS4/4
mistralaidevstral-smallS4/4
mistralaiministral-14b-2512S4/4
mistralaiministral-3b-2512S4/4
mistralaiministral-8b-2512S4/4
mistralaimistral-largeS4/4
mistralaimistral-large-2407S4/4
mistralaimistral-large-2411S4/4
mistralaimistral-large-2512S4/4
mistralaimistral-medium-3S4/4
mistralaimistral-medium-3.1S4/4
mistralaimistral-nemoS4/4
mistralaimistral-sabaS4/4
mistralaimistral-small-2603S4/4
mistralaimistral-small-3.2-24b-instructS4/4
mistralaimistral-small-creativeS4/4
mistralaimixtral-8x22b-instructS4/4
mistralaipixtral-large-2411S4/4
mistralaivoxtral-small-24b-2507S4/4
moonshotaikimi-k2S4/4
moonshotaikimi-k2-0905S4/4
moonshotaikimi-k2.5S4/4
nex-agideepseek-v3.1-nex-n1S4/4
nvidiallama-3.1-nemotron-70b-instructS4/4
nvidiallama-3.3-nemotron-super-49b-v1.5S4/4
nvidianemotron-3-nano-30b-a3bS4/4
nvidianemotron-nano-9b-v2S4/4
openaigpt-3.5-turboS4/4
openaigpt-3.5-turbo-0613S4/4
openaigpt-3.5-turbo-16kS4/4
openaigpt-4S4/4
openaigpt-4-turboS4/4
openaigpt-4.1S4/4
openaigpt-4.1-miniS4/4
openaigpt-4.1-nanoS4/4
openaigpt-4oS4/4
openaigpt-4o-2024-05-13S4/4
openaigpt-4o-2024-08-06S4/4
openaigpt-4o-2024-11-20S4/4
openaigpt-4o-miniS4/4
openaigpt-4o-mini-2024-07-18S4/4
openaigpt-5S4/4
openaigpt-5-miniS4/4
openaigpt-5-nanoS4/4
openaigpt-5-proS4/4
openaigpt-5.1S4/4
openaigpt-5.1-chatS4/4
openaigpt-5.2S4/4
openaigpt-5.2-chatS4/4
openaigpt-5.3-chatS4/4
openaigpt-5.4S4/4
openaigpt-5.4-miniS4/4
openaigpt-5.4-nanoS4/4
openaigpt-oss-120bS4/4
openaigpt-oss-20bS4/4
openaio1S4/4
openaio3S4/4
openaio3-miniS4/4
openaio3-mini-highS4/4
openaio3-proS4/4
openaio4-miniS4/4
openaio4-mini-highS4/4
openrouterautoS4/4
openrouterfreeS4/4
prime-intellectintellect-3S4/4
qwenqwen-2.5-72b-instructS4/4
qwenqwen-2.5-7b-instructS4/4
qwenqwen-maxS4/4
qwenqwen-plusS4/4
qwenqwen-plus-2025-07-28S4/4
qwenqwen-turboS4/4
qwenqwen3-14bS4/4
qwenqwen3-235b-a22bS4/4
qwenqwen3-235b-a22b-2507S4/4
qwenqwen3-30b-a3bS4/4
qwenqwen3-30b-a3b-instruct-2507S4/4
qwenqwen3-32bS4/4
qwenqwen3-8bS4/4
qwenqwen3-coderS4/4
qwenqwen3-coder-30b-a3b-instructS4/4
qwenqwen3-coder-flashS4/4
qwenqwen3-coder-nextS4/4
qwenqwen3-coder-plusS4/4
qwenqwen3-maxS4/4
qwenqwen3-next-80b-a3b-instructS4/4
qwenqwen3.5-122b-a10bS4/4
qwenqwen3.5-27bS4/4
qwenqwen3.5-35b-a3bS4/4
qwenqwen3.5-397b-a17bS4/4
qwenqwen3.5-9bS4/4
qwenqwen3.5-flash-02-23S4/4
qwenqwen3.5-plus-02-15S4/4
qwenqwq-32bS4/4
relacerelace-searchS4/4
stepfunstep-3.5-flashS4/4
tngtechdeepseek-r1t2-chimeraS4/4
upstagesolar-pro-3S4/4
x-aigrok-3S4/4
x-aigrok-3-betaS4/4
x-aigrok-3-miniS4/4
x-aigrok-3-mini-betaS4/4
x-aigrok-4S4/4
x-aigrok-4-fastS4/4
x-aigrok-4.1-fastS4/4
x-aigrok-4.20S4/4
x-aigrok-code-fast-1S4/4
xiaomimimo-v2-flashS4/4
xiaomimimo-v2-omniS4/4
xiaomimimo-v2-proS4/4
z-aiglm-4-32bS4/4
z-aiglm-4.5S4/4
z-aiglm-4.5-airS4/4
z-aiglm-4.5vS4/4
z-aiglm-4.6S4/4
z-aiglm-4.6vS4/4
z-aiglm-4.7S4/4
z-aiglm-4.7-flashS4/4
z-aiglm-5S4/4
z-aiglm-5-turboS4/4
z-aiglm-5v-turboS4/4
deepseekdeepseek-r1-0528A3/4
rekaaireka-edgeA3/4
openaigpt-oss-safeguard-20bB2/4
ai21jamba-large-1.7C1/4
arcee-aitrinity-miniC1/4
meituanlongcat-flash-chatC1/4
mistralaimixtral-8x7b-instructF0/4

Methodology

Every model is tested through OpenRouter using the OpenAI-compatible API format. Each model runs 4 real tool-calling scenarios against a live AgentNode Runtime instance with actual registered tools.

A scenario passes when the model calls all expected tools with correct arguments and produces a coherent response. No mocking, no simulated tool calls. The model must generate valid function calls that the runtime can execute.

Models that return API errors (rate limits, unavailable, etc.) are excluded from the matrix and not counted in any tier.

Works with your model

96% of all tested models achieve perfect compatibility. Pick any model from your provider and start building.