OpenWeightsTerminal
Inference

The right model. A clear price.

Compare reference prices with what the network can serve now.

Computers online
0
Model aliases
0
Reported tok/s · sum
0

What these numbers count

One computer can share several models. Reported throughput counts its highest model speed once; it is not simultaneous capacity for every model.

USD per 1 million tokens · 40 families / endpointsSelect up to 3 to compare

40 families / endpoints

Reference only (40)

Ling 3.0 Flash

Reference only
Reference · per 1M tokens
Input
$0.021
Output
$0.063
Model & price details
Catalog reference
inclusionAI
Reference input / output
$0.021 / $0.063
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Nemotron 3.5 Lightning

Reference only
Reference · per 1M tokens
Input
$0.080
Output
$0.20
Model & price details
Catalog reference
NVIDIA
Reference input / output
$0.080 / $0.20
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Gemma 4 26B

Reference only
Reference · per 1M tokens
Input
$0.090
Output
$0.30
Model & price details
Catalog reference
Google
Reference input / output
$0.090 / $0.30
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Gemma 4 31B

Reference only
Reference · per 1M tokens
Input
$0.090
Output
$0.34
Model & price details
Catalog reference
TrevorJS
Reference input / output
$0.090 / $0.34
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.8 Flash

Reference only
Reference · per 1M tokens
Input
$0.15
Output
$0.47
Model & price details
Catalog reference
Qwen
Reference input / output
$0.15 / $0.47
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

GLM 5.3 Flash

Reference only
Reference · per 1M tokens
Input
$0.15
Output
$0.50
Model & price details
Catalog reference
Zhipu
Reference input / output
$0.15 / $0.50
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek v4.1 Flash

Reference only
Reference · per 1M tokens
Input
$0.15
Output
$0.60
Model & price details
Catalog reference
DeepSeek
Reference input / output
$0.15 / $0.60
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.6 35B-A3B

Reference only
Reference · per 1M tokens
Input
$0.15
Output
$1.00
Model & price details
Catalog reference
Qwen
Reference input / output
$0.15 / $1.00
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.8 27B

Reference only
Reference · per 1M tokens
Input
$0.42
Output
$3.00
Model & price details
Catalog reference
Qwen
Reference input / output
$0.42 / $3.00
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

GLM 5.2

Reference only
Reference · per 1M tokens
Input
$0.65
Output
$2.04
Model & price details
Catalog reference
Zhipu
Reference input / output
$0.65 / $2.04
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

GLM 5.3

Reference only
Reference · per 1M tokens
Input
$1.40
Output
$4.40
Model & price details
Catalog reference
Zhipu
Reference input / output
$1.40 / $4.40
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

AliceAI-Foundation-80B-A3B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Yandex
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

CYBER-FROST 3.8

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Blackfrost-AI
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek R1 · 14B Uncensored

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek R1 · 14B Uncensored
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek R1 · 7B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek R1 · 7B
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek R1 · 7B Uncensored

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek R1 · 7B Uncensored
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek V4 Flash 0731

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek V4 Flash 0731
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek V4 Flash 0731 · GGUF

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek V4 Flash 0731 · GGUF
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek V4.1 Flash · ds4 Q2

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek V4.1 Flash · ds4 Q2
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

DeepSeek V4.1 Flash Abliterated · ds4 Q2

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
DeepSeek V4.1 Flash Abliterated · ds4 Q2
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Dolphin 3.0 · Llama 3.1 8B

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Dolphin 3.0 · Llama 3.1 8B
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Gemma-4 12B OBLITERATED

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
OBLITERATUS
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Huihui DeepSeek V4 Flash 0731 Abliterated · GGUF

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Huihui DeepSeek V4 Flash 0731 Abliterated · GGUF
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Kimi K3

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Moonshot
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

MiMo-V2.6-Distill-Qwen-9B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Xiaomi MiMo
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

MiMo-V2.6-Flash

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Xiaomi MiMo
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

MiniMax-M2.5

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
MiniMax
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Nemotron 3.5 Bolt 74B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
NVIDIA
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Nex-N2.5-mini

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
community
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Ornith

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
ornith-ai
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen-Image-2.1

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Qwen
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen-Image-2.1 Uncensored

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
abenzerps
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3-8B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Qwen
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.5 2B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Qwen
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.5-4B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Qwen
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.5-9B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Qwen
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Qwen3.8-9B Heretic

Uncensored
Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Noobito45
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

Ternary Bonsai 2 27B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
PrismML
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured
U

Underdog Saluki 27B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
Underdog AI
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured
W

WhiteRabbitNeo 2.5 · Cyber 7B

Reference only
Reference · per 1M tokens
Input
—
Output
—
Model & price details
Catalog reference
WhiteRabbitNeo 2.5 · Cyber 7B
Reference input / output
— / —
Snapshot · 2026-09-24
Speed · tok/s
— · Not measured

OpenRouter reference snapshot Dated 2026-09-24, not a live provider quote. Host rates apply to input plus output tokens. Priority changes points, not price. Endpoint tests do not attest weights. Self-reported speed is not a throughput guarantee.

For hosts

Start with a supported build.

How contribution points work

Open-weight and uncensored builds are prioritized. Catalog inclusion permits review; it does not prove a host’s weights. There is no automatic price premium. Read each license before hosting.

Boost requires a linked account, desk review, a test in the last 24h and eligible paid usage.

Dolphin 3.0 · Llama 3.1 8B

dphn

Accepted build
BF16 source weights
Memory guidance
~16 GB weights only; add runtime and context-cache headroom
Source & license
License
Llama 3.1 Community License; review terms before hosting

Qwen3.5 · 4B

Qwen

Accepted build
BF16 source weights
Memory guidance
~8 GB weights only; add runtime and context-cache headroom
Source & license
License
Apache 2.0

Another build? Connect it normally, run an endpoint check and contact the desk for catalog review. Unlisted builds do not earn contribution points.