G9v3-3B
G9v3-3B by AI9Stars is a dense causal language model for local use. Features 131K context, Think/No Think modes, targeting coding, tool-use, and reasoning.
Model Overview
Capabilities, design details, and architectural traits
G9v3-3B - Compact Dense Model for Local Deployment
G9v3-3B is a dense causal language model from AI9Stars, built specifically for local deployment and resource-constrained scenarios. It uses a standard LlamaForCausalLM architecture and targets everyday assistant use, coding, tool-use workflows, and reasoning tasks where a compact footprint matters.
A defining feature is its Think / No Think toggle, controlled via the enable_thinking parameter in the chat template. In Think mode, recommended sampling uses temperature=0.9, top_p=0.95. In No Think mode, recommended sampling shifts to temperature=0.7, top_p=0.95. This lets users trade between extended reasoning and faster direct responses within the same model.
| Trait | Detail |
|---|---|
| Architecture | Standard LlamaForCausalLM, dense (non-MoE) |
| Context length | 131,072 tokens |
| Think / No Think toggle | enable_thinking parameter switches between reasoning and direct-response modes |
| Deployment focus | Local and resource-constrained environments |
| Target workloads | Coding, tool-use workflows, reasoning, everyday assistant use |
| Precision | BF16 |
| License | Apache-2.0 |
Inference Ecosystem
The model card provides quickstart guides for vLLM, SGLang, and Transformers, reinforcing its local-first positioning. Quantized versions are available for llama.cpp, LM Studio, Jan, and Ollama, and the model tree includes community adapters and finetunes.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | AI9Stars | Anthropic | Anthropic |
| Release Date | July 23, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 131K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | Free Best Input Pricing | $4 | $2 |
| Output Pricing | Free Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 10.8 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | 9.9 | - | - |
| Agentic Index | 14.0 | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from AI9Stars
Other models by AI9Stars
Top AI Models
Leading alternatives by intelligence score
