Catalogue

every model we serve, and what it really costs

One OpenAI-compatible endpoint, open weights only, priced per million tokens with no minimum and no subscription. Each model reaches several classes of GPU through different quantizations, so a 6 GB card and a 24 GB card both have work to do — and hosts keep 80% of what their hardware earns.

what your GPU can serve

your card	models	for example
a 6 GB card	3 models	Gemma 4 E4B Instruct, Qwen2.5 1.5B Instruct, Qwen2.5 0.5B Instruct
8 GB	3 models	Gemma 4 E4B Instruct, Qwen2.5 1.5B Instruct, Qwen2.5 0.5B Instruct
10 GB	5 models	Qwen3.8 27B Instruct, Gemma 4 12B Instruct, Gemma 4 E4B Instruct, Qwen2.5 1.5B Instruct
12 GB	7 models	Qwen3.8 27B Instruct, GLM-4.7-Flash, Qwen3 Coder 30B-A3B, Gemma 4 12B Instruct
16 GB	9 models	Qwen3.8 27B Instruct, Qwen3.6 35B-A3B, GLM-4.7-Flash, Qwen3 Coder 30B-A3B
24 GB	10 models	Qwen3.8 27B Instruct, Qwen3.6 35B-A3B, GLM-4.7-Flash, Qwen3 Coder 30B-A3B

Numbers are the minimum VRAM a build needs. Install mahout, run mahout models, and it tells you exactly what your own card can serve.

Qwen3.8 27B Instruct

qwen3.8-27b

2026 flagship open model for a single card: strong reasoning and code, 262K context, and no EU-sovereign listing anywhere else.

chat code reasoning long-context multilingual tool-use vision

size: 27B dense
max context: 256K tokens
licence: Apache-2.0
price: €0.25 per million input tokens · €0.75 per million output tokens
smallest card: from 9 GB VRAM

9 build(s) — one per VRAM class

quantization	download	needs	context	notes
`UD-Q5_K_M`	18.4 GB	23 GB VRAM	32K	reference quality
`UD-Q4_K_M`	15.3 GB	20 GB VRAM	32K	reference quality
`UD-IQ4_XS`	13.3 GB	18 GB VRAM	32K	balanced
`UD-Q3_K_XL`	12.2 GB	16 GB VRAM	32K	balanced
`UD-IQ3_S`	11.2 GB	15 GB VRAM	32K	balanced
`UD-IQ3_XXS`	10.2 GB	14 GB VRAM	32K	compact — fits smaller cards
`UD-Q2_K_XL`	9.2 GB	12 GB VRAM	32K	compact — fits smaller cards
`UD-IQ2_S`	7.8 GB	11 GB VRAM	16K	compact — fits smaller cards
`UD-IQ2_XXS`	6.8 GB	9 GB VRAM	16K	compact — fits smaller cards

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Qwen3.6 35B-A3B

qwen3.6-35b-a3b

Fast enough to delegate to: 3B active parameters decode at small-model speed while the full 35B answers, which is what makes it the one to hand sub-agent work to. Sees images, 262K context.

chat code reasoning long-context multilingual tool-use vision

size: 35B MoE, 3B active
max context: 256K tokens
licence: Apache-2.0
price: €0.18 per million input tokens · €0.55 per million output tokens
smallest card: from 13 GB VRAM

8 build(s) — one per VRAM class

quantization	download	needs	context	notes
`UD-IQ4_NL_XL`	18.2 GB	23 GB VRAM	32K	reference quality
`UD-IQ4_XS`	16.5 GB	21 GB VRAM	32K	reference quality
`UD-Q3_K_XL`	15.7 GB	20 GB VRAM	32K	balanced
`UD-IQ3_S`	12.7 GB	17 GB VRAM	32K	balanced
`UD-IQ3_XXS`	12.3 GB	16 GB VRAM	32K	balanced
`UD-Q2_K_XL`	11.4 GB	15 GB VRAM	32K	compact — fits smaller cards
`UD-IQ2_XXS`	10.0 GB	14 GB VRAM	32K	compact — fits smaller cards
`UD-IQ1_M`	9.4 GB	13 GB VRAM	16K	compact — fits smaller cards

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

GLM-4.7-Flash

glm-4.7-flash

MIT licensed end to end — the most permissive terms in the catalogue, and the answer for buyers whose legal team reads the licence. Strong general reasoning from a second vendor.

chat code reasoning long-context multilingual tool-use

size: 31B MoE
max context: 198K tokens
licence: MIT
price: €0.16 per million input tokens · €0.5 per million output tokens
smallest card: from 11 GB VRAM

9 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_1`	17.7 GB	23 GB VRAM	32K	reference quality
`UD-Q4_K_XL`	16.3 GB	21 GB VRAM	32K	reference quality
`IQ4_XS`	15.2 GB	20 GB VRAM	32K	balanced
`UD-Q3_K_XL`	12.8 GB	17 GB VRAM	32K	balanced
`UD-IQ3_XXS`	12.0 GB	16 GB VRAM	32K	balanced
`UD-Q2_K_XL`	11.1 GB	15 GB VRAM	32K	compact — fits smaller cards
`UD-IQ2_XXS`	9.8 GB	14 GB VRAM	32K	compact — fits smaller cards
`UD-IQ1_S`	8.6 GB	12 GB VRAM	16K	compact — fits smaller cards
`UD-TQ1_0`	7.8 GB	11 GB VRAM	16K	compact — fits smaller cards

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Qwen3 Coder 30B-A3B

qwen3-coder-30b-a3b

Built for code and agentic tool use: repository-scale context at 262K, and 3B active parameters so it answers at conversational speed.

chat code reasoning long-context tool-use

size: 30B MoE, 3B active
max context: 256K tokens
licence: Apache-2.0
price: €0.2 per million input tokens · €0.6 per million output tokens
smallest card: from 11 GB VRAM

9 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_1`	17.9 GB	23 GB VRAM	32K	reference quality
`UD-Q4_K_XL`	16.5 GB	21 GB VRAM	32K	reference quality
`IQ4_XS`	15.3 GB	20 GB VRAM	32K	balanced
`UD-Q3_K_XL`	12.9 GB	17 GB VRAM	32K	balanced
`UD-IQ3_XXS`	12.0 GB	16 GB VRAM	32K	balanced
`UD-Q2_K_XL`	11.0 GB	15 GB VRAM	32K	compact — fits smaller cards
`UD-IQ2_XXS`	9.6 GB	14 GB VRAM	32K	compact — fits smaller cards
`UD-IQ1_M`	9.0 GB	13 GB VRAM	16K	compact — fits smaller cards
`UD-TQ1_0`	7.5 GB	11 GB VRAM	16K	compact — fits smaller cards

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Gemma 4 26B-A4B Instruct

gemma-4-26b-a4b

Mixture-of-experts: 26B of knowledge at the speed of a 4B model. Quantization-aware weights published by Google, so the 4-bit build loses far less than a post-hoc quantization.

chat reasoning multilingual long-context

size: 26B MoE, 4B active
max context: 128K tokens
licence: Gemma Terms of Use
price: €0.15 per million input tokens · €0.45 per million output tokens
smallest card: from 17 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_0 (QAT)`	13.4 GB	17 GB VRAM	32K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

gpt-oss 20B

gpt-oss-20b

Open-weight 20B tuned for tool use and structured output. Natively MXFP4, so 4-bit costs it almost nothing.

chat code tool-use reasoning

size: 20B MoE
max context: 128K tokens
licence: Apache-2.0
price: €0.12 per million input tokens · €0.4 per million output tokens
smallest card: from 14 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_K_M`	10.8 GB	14 GB VRAM	32K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Gemma 4 12B Instruct

gemma-4-12b

The 8 GB workhorse: a capable general assistant on cards most gamers already own, with Google's quantization-aware 4-bit weights.

chat reasoning multilingual

size: 12B dense
max context: 128K tokens
licence: Gemma Terms of Use
price: €0.08 per million input tokens · €0.25 per million output tokens
smallest card: from 10 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_0 (QAT)`	6.5 GB	10 GB VRAM	32K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Gemma 4 E4B Instruct

gemma-4-e4b

Built for small cards and laptops: 6 GB of VRAM is enough to earn.

chat multilingual

size: 8B MatFormer, 4B effective
max context: 128K tokens
licence: Gemma Terms of Use
price: €0.05 per million input tokens · €0.15 per million output tokens
smallest card: from 6 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_0 (QAT)`	4.8 GB	6 GB VRAM	16K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Qwen2.5 1.5B Instruct

qwen2.5-1.5b-instruct

Tiny, fast, runs anywhere — the enrollment smoke test and the model behind starter work for a freshly linked machine.

chat

size: 1.5B dense
max context: 32K tokens
licence: Apache-2.0
price: €0.05 per million input tokens · €0.15 per million output tokens
smallest card: from 3 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_K_M`	1.0 GB	3 GB VRAM	32K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

Qwen2.5 0.5B Instruct

qwen2.5-0.5b-instruct

Smallest catalog entry: CPU-only fallback and multi-model swap tests.

chat

size: 0.5B dense
max context: 32K tokens
licence: Apache-2.0
price: €0.02 per million input tokens · €0.06 per million output tokens
smallest card: from 2 GB VRAM

1 build(s) — one per VRAM class

quantization	download	needs	context	notes
`Q4_K_M`	0.5 GB	2 GB VRAM	32K	reference quality

Every build is pinned to an exact upstream revision and verified against its SHA-256 before it runs — on your machine, if you host, and on the machine serving you, if you build.

use them, or serve them

Point your SDK's base URL at https://app.elephantpool.ai/api/gateway/v1 and pass the model id — or connect a GPU and get paid for every token it produces.

Get an API key Serve with your GPU How mahout works Knowledge base