The Known Good Updated 26 Oct 2026 Subscribe
Providers· Inference host· 12 tracked models

InferenceNet — pricing & performance

Compare models All providers Data sources
Models served
12
of 1,234 tracked models
Mean output speed
106
tokens per second, averaged over
12 measured endpoints
Mean latency
0.6s
time to first token, averaged over
12 measured endpoints
Cheapest blended
$0.060
per 1M tokens at a 3:1
input:output blend
Output speed on InferenceNet

Output tokens per second for all 12 models on InferenceNet with published throughput telemetry, the most recent figure OpenRouter publishes for each host. Higher is better. We do not run these measurements ourselves.

⇩
219
201
118
112
101
100
94
87
67
63
63
44
The Known Good
12 models on InferenceNet
Full leaderboard →
Model Creator Input $/M Output $/M Blended Speed TTFT Context
DeepSeek: DeepSeek V4.1 Flash DeepSeek $0.10 $0.60 $0.23 118 0.3s 1.04M
DeepSeek: DeepSeek V4.1 Flash 1017 DeepSeek $0.080 $0.55 $0.20 101 0.4s 1.04M
Inference.net: Schematron V2 Small inference-net $0.050 $0.23 $0.095 219 0.2s 128k
Inference.net: Schematron V2 Turbo inference-net $0.030 $0.15 $0.060 63 0.6s 128k
MiniMax: MiniMax M3.1 Flash MiniMax $0.15 $0.60 $0.26 67 0.6s 1.00M
MoonshotAI: Kimi K3 Moonshot AI $0.24 $10.80 $2.88 100 0.5s 1.05M
StepFun: Step 5 Preview stepfun $1.00 $2.70 $1.43 201 0.5s 1.05M
Tencent: Hy4 preview Tencent $0.75 $2.25 $1.12 44 1.9s 1.05M
Xiaomi: MiMo-V2.6-Flash Xiaomi $0.12 $0.28 $0.16 63 0.9s 1.05M
Z.ai: GLM 5.2 Z.ai $1.30 $4.40 $2.08 87 0.7s 1.05M
Z.ai: GLM 5.3 Z.ai $0.080 $3.60 $0.96 112 0.7s 1.05M
Z.ai: GLM 5.3 Flash Z.ai $0.12 $0.39 $0.19 94 0.3s 1.05M
The Known Good
Source. Prices, endpoint availability and throughput for InferenceNet are ingested from OpenRouter; model metadata is enriched from models.dev. A blank cell means the figure was not published for that endpoint — it never means zero. Blended price is (3 × input + output) ÷ 4, the same 3:1 blend used everywhere on this site; see methodology. Quality scores are not provider-specific: a model scores the same wherever it is hosted, so quality lives on the model pages.