The Known Good Updated 24 Oct 2026 Subscribe
Providers· Inference host· 14 tracked models

Darkbloom — pricing & performance

Compare models All providers Data sources
Models served
14
of 1,302 tracked models
Mean output speed
26
tokens per second, averaged over
14 measured endpoints
Mean latency
2.9s
time to first token, averaged over
14 measured endpoints
Cheapest blended
$0.000
per 1M tokens at a 3:1
input:output blend
Output speed on Darkbloom

Output tokens per second for all 14 models on Darkbloom with published throughput telemetry, the most recent figure OpenRouter publishes for each host. Higher is better. We do not run these measurements ourselves.

⇩
70
40
39
30
29
29
26
23
20
14
12
11
The Known Good
14 models on Darkbloom
Full leaderboard →
Model Creator Input $/M Output $/M Blended Speed TTFT Context
Google: Gemma 4 26B A4B Google $0.042 $0.22 $0.086 40 0.6s 131k
Google: Gemma 4 26B A4B (free) Google $0.000 $0.000 $0.000 12 5.5s 131k
NVIDIA: Nemotron 3.5 Lightning NVIDIA $0.025 $0.12 $0.049 70 1.2s 262k
NVIDIA: Nemotron 3.5 Super VL NVIDIA $0.060 $0.35 $0.13 11 6.3s 262k
OpenAI: gpt-oss-20b OpenAI $0.018 $0.090 $0.036 39 1.4s 131k
OpenAI: gpt-oss-20b (free) OpenAI $0.000 $0.000 $0.000 20 3.7s 131k
PrismML: Ternary Bonsai 2 27B prism-ml $0.075 $0.50 $0.18 26 2.8s 262k
Qwen: Qwen3 VL 30B A3B Instruct Alibaba $0.090 $0.40 $0.17 6 1.7s 131k
Qwen: Qwen3.5-35B-A3B Alibaba $0.065 $0.75 $0.24 30 3.8s 262k
Qwen: Qwen3.5-9B Alibaba $0.080 $0.13 $0.092 29 1.0s 262k
Qwen: Qwen3.6 35B A3B Alibaba $0.050 $0.70 $0.21 23 7.0s 262k
Qwen: Qwen3.8 27B Alibaba $0.035 $2.20 $0.58 29 2.1s 262k
Xiaomi: MiMo-V2.6-Flash Xiaomi $0.10 $0.28 $0.14 14 2.8s 1.05M
Z.ai: GLM 5.3 Flash Z.ai $0.065 $0.18 $0.094 11 1.2s 262k
The Known Good
Source. Prices, endpoint availability and throughput for Darkbloom 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.