Skip to content
RankShark

Ternary Bonsai 2 9B

Ternary Bonsai 2 9B is PrismML’s compact, open-weight language model for conversational AI, reasoning, and tool use.

ModelOpen sourceby PrismMLsince 2026
62.3

SharkScore

#9 in Models

Why it scores 62.3

Ternary Bonsai 2 9B has a SharkScore of 62.3/100, ranked #9 of 608 in Models (updated 2026-11-19).

Each part compares Ternary Bonsai 2 9B with other listings in Models. Evidence covers 87% of the score's weight.

Adoption
37 · 30%
User reviews
no data · 25%
Maintenance
93 · 20%
Trust & transparency
80 · 15%
Momentum
no data · 10%

Paying never changes this score. How ranking works · method v2026-11-18.2

RankShark review of Ternary Bonsai 2 9B

Automated evaluation by RankShark · 18 Nov 2026 · method v1.0 · How we review

We were not paid for this review.

As of 2026-11-18, Ternary Bonsai 2 9B is an open-weight vision-language model with a 262,144-token context; OpenRouter lists usage pricing of $0.08 per million input tokens and $0.13 per million output tokens. (E1; E2)

PrismML’s approximately 9B-parameter vision-language model uses ternary weights stored in GGUF formats. It accepts text and images, and is available through OpenRouter or for local use with supported runtimes. (E1; E2)

Best for

  • People looking to run a compact-weight model locally with llama.cpp on CUDA, Metal, or CPU. (E1) · model card
  • Applications that need text and image input with a listed 262,144-token context. (E1; E2) · model card, OpenRouter

Not for

  • Users who need PTQ1_0 or PQ2_0 files to run on upstream llama.cpp; the model page says those formats require PrismML’s fork. (E1) · model card

Key features

  • Ternary GGUF formats. The vendor says the same ternary language weights are available as PTQ1_0, PQ2_0, and Q2_0 files, ranging from 2.00 GB to 2.56 GB. (E1) · model card
  • Image input. The vendor says image input uses a separate vision projector file, available in Q8_0 or BF16. (E1) · model card
  • Long context. The vendor lists a 262,144-token context inherited from the Qwen3.5-9B backbone. (E1) · model card
  • Speculative decoding files. The vendor provides optional MTP and DFlash2 drafter files; both require the PrismML llama.cpp fork. (E1) · model card
  • Local and hosted access. The vendor documents an OpenAI-compatible llama.cpp server; OpenRouter lists chat completions, completions, messages, and responses endpoints. (E1; E2) · model card, OpenRouter

Pros

  • The vendor lists three GGUF packings, with the smallest language-model file at 2.00 GB. (E1) · model card
  • The vendor documents both text-only and image-plus-text use. (E1) · model card
  • OpenRouter lists a 262,144-token context and a maximum completion length of 65,536 tokens. (E2) · OpenRouter

Cons

  • The vendor says upstream llama.cpp does not support PTQ1_0 or PQ2_0; those formats require the PrismML fork. (E1) · model card
  • The vendor warns that the BF16 vision projector can crash llama-server when several image requests run at once in the current fork build. (E1) · model card
  • No free plan or free trial stated on the pages we checked.
  • Whether customer data is used for model training is not stated on the pages we checked.

Pricing

PlanPriceLimits
OpenRouter API usage$0.08 per million input tokens; $0.13 per million output tokensOpenRouter lists a 262,144-token context and a 65,536-token maximum completion. (E2)

Checked 18 Nov 2026

Free plan
Not stated (checked 18 Nov 2026)
Free trial
Not stated (checked 18 Nov 2026)
Trains on customer data
Not stated (checked 18 Nov 2026)
Data retention
Not stated (checked 18 Nov 2026)
Data regions
Not stated (checked 18 Nov 2026)
Data processing agreement
Not stated (checked 18 Nov 2026)
AI models used
Ternary Bonsai 2 9B; the model card says it is derived from Qwen3.5-9B. · model card, OpenRouter
API
OpenRouter lists chat completions, completions, messages, and responses endpoints. The model card also documents a local OpenAI-compatible llama.cpp server. · model card, OpenRouter
SDKs
Not stated (checked 18 Nov 2026)
Webhooks
Not stated (checked 18 Nov 2026)
Platforms
The model card lists llama.cpp with CUDA, Metal, and CPU backends, and also documents an MLX build for Apple Silicon. · model card
Security claims
No security certification stated (checked 18 Nov 2026)

What we measured

Context window (OpenRouter listing) · now
262,144 tokens
Price per million tokens (OpenRouter listing) · now
$0.08 in, $0.13 out
Package downloads · 30 days
1,548,119
How we tested · sources
  • Hourly uptime and response-time checks of the product's URL.
  • Repository, release and download data from GitHub, npm, PyPI and Hugging Face where they exist.
  • We did not sign up, pay, or use the product as a customer.
  • 18 Nov 2026: Plans or prices changed on the pricing page.

Something wrong? Report an error with a public link that shows it; corrections re-run the review.

Top alternatives to Ternary Bonsai 2 9B

See all
  • 67.1

    Qwen3.8 27B is an open-weight dense vision-language model from Qwen.

    ModelOpen source
  • 65.5

    Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per t

    ModelOpen source
  • 64.7

    Qwen3.8 Flash is a multimodal reasoning model from Alibaba.

    ModelOpen source
  • 64.3

    Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.

    ModelOpen source