The Known Good Updated 29 Jul 2026
Providers· Inference host· 19 tracked models

Phala — pricing & performance

Compare models All providers Data sources
Models served
19
of 947 tracked models
Mean output speed
44
tokens per second, averaged over
15 measured endpoints
Mean latency
1.6s
time to first token, averaged over
15 measured endpoints
Cheapest blended
$0.055
per 1M tokens at a 3:1
input:output blend
Output speed on Phala

Median output tokens per second for the 15 of 19 models on Phala with published throughput telemetry from OpenRouter. The full table below lists all 19, measured or not. Higher is better. We do not run these measurements ourselves.

⇩
93
82
73
59
58
50
42
42
36
27
24
23
The Known Good
19 models on Phala
Full leaderboard →
Model Creator Input $/M Output $/M Blended Speed TTFT Context
DeepSeek: DeepSeek V3.2 DeepSeek $1.00 $1.00 $1.00 13 3.9s 164k
Google: Gemma 3 27B Google $0.15 $0.46 $0.23 20 1.7s 262k
Google: Gemma 4 31B Google $0.15 $0.46 $0.23 18 1.3s 262k
MiniMax: MiniMax M2.5 MiniMax $0.20 $1.38 $0.49 23 1.6s 197k
MoonshotAI: Kimi K2.5 Moonshot AI $0.60 $3.00 $1.20 27 1.3s 262k
MoonshotAI: Kimi K2.6 Moonshot AI $1.09 $4.60 $1.97 42 1.5s 262k
OpenAI: gpt-oss-120b OpenAI $0.15 $0.60 $0.26 82 0.9s 131k
OpenAI: gpt-oss-20b OpenAI $0.040 $0.15 $0.068 58 0.4s 131k
Qwen: Qwen2.5 7B Instruct Alibaba $0.040 $0.10 $0.055 59 0.3s 33k
Qwen: Qwen3 30B A3B Instruct 2507 Alibaba $0.15 $0.55 $0.25 93 1.2s 262k
Qwen: Qwen3 VL 30B A3B Instruct Alibaba $0.20 $0.70 $0.33 — — 128k
Qwen: Qwen3.5 397B A17B Alibaba $0.55 $3.50 $1.29 24 2.0s 262k
Qwen: Qwen3.5-27B Alibaba $0.30 $2.40 $0.82 — — 262k
Qwen: Qwen3.6 27B Alibaba $0.32 $2.70 $0.92 — — 262k
Qwen: Qwen3.6 35B A3B Alibaba $0.20 $1.27 $0.47 73 0.9s 262k
Z.ai: GLM 4.7 Z.ai $0.85 $3.30 $1.46 36 3.8s 131k
Z.ai: GLM 5 Z.ai $1.20 $3.50 $1.77 — — 203k
Z.ai: GLM 5.1 Z.ai $1.21 $4.20 $1.96 42 2.0s 203k
Z.ai: GLM 5.2 Z.ai $1.40 $4.40 $2.15 50 1.4s 1.05M
The Known Good
Source. Prices, endpoint availability and throughput for Phala 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.