Arcee AI releases Trinity Large Thinking, open-source reasoning model with 262K context window
Arcee AI has released Trinity Large Thinking, an open-source reasoning model featuring a 262,144 token context window. The model is priced at $0.25 per million input tokens and $0.90 per million output tokens, with free access available through OpenRouter for the first five days.
Trinity Large Thinking — Quick Specs
Arcee AI has released Trinity Large Thinking, an open-source reasoning model designed for complex problem-solving tasks. The model launched on April 1, 2026, and is available through OpenRouter with immediate access.
Specifications
Trinity Large Thinking offers a 262,144 token context window, enabling processing of substantial documents and extended conversations. The model supports OpenRouter's reasoning-enabled architecture, allowing users to access the model's step-by-step thinking process through the reasoning_details array in API responses.
Pricing
Input tokens cost $0.25 per million, while output tokens cost $0.90 per million. Arcee AI is offering free access through OpenRouter for the first five days following launch, allowing developers to evaluate the model at no cost.
Performance and Capabilities
According to Arcee AI, the model demonstrates strong performance on PinchBench, agentic workloads, and reasoning tasks. The reasoning capabilities enable transparent access to the model's internal decision-making process, allowing users to preserve and continue reasoning chains across multi-turn conversations.
Technical Details
Trinity Large Thinking is classified as a reasoning model and is available as open-source weights. OpenRouter routes requests across multiple providers with automatic fallback capability to maximize uptime. The platform normalizes requests and responses across different provider infrastructures.
Developers can integrate the model using standard OpenRouter API endpoints, with optional framework support through third-party SDKs. The reasoning parameter enables thinking mode in requests, while continuing conversations requires preservation of complete reasoning_details to maintain context continuity.
What this means
The release marks Arcee AI's entry into the reasoning model category with a competitive context window size. At $0.25/$0.90 pricing, Trinity Large Thinking positions itself in the mid-range of open-source reasoning models. The five-day free trial strategy reduces friction for adoption testing. The transparent reasoning process differentiates it from black-box alternatives, though real-world performance claims require independent verification against established benchmarks.
Related Articles
Google launches Nano Banana 2 Lite image model at 4 seconds per image, $0.04 per 1,000 generations
Google released Nano Banana 2 Lite, an image generation model that produces images in four seconds at under four cents per thousand images. The model prioritizes speed and cost over quality, targeting developers building high-volume image pipelines.
Google launches Gemini 3.1 Flash Lite Image with 4-second generation time, $0.25 per 1M input tokens
Google has released Gemini 3.1 Flash Lite Image, a text-to-image model that generates 1K resolution images in approximately 4 seconds — 2.7× faster than Gemini 3.1 Flash Image. The model is priced at $0.25 per 1M input tokens and $1.50 per 1M output tokens, with a 66K context window and knowledge cutoff of January 2025.
DeepReinforce Releases Ornith-1.0, Open-Source Agentic Coding Model in 9B to 397B Sizes
DeepReinforce has released Ornith-1.0, an MIT-licensed model designed for agentic coding tasks with variants ranging from 9B to 397B parameters. Built on top of Apache 2.0-licensed Gemma 4 and Qwen 3.5 base models, the company claims it achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks.
DeepSeek Releases V4 Models: 1M Context Window, 90% Less KV Cache Than V3
DeepSeek has released two new MoE models: DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated). Both models support a one million token context window and use a hybrid attention architecture that requires only 27% of single-token inference FLOPs and 10% of KV cache compared to DeepSeek-V3.2.
Comments
Loading...