LiquidAI: LFM3-1.7B (free)

ModelsAtlas catalogs LiquidAI's LFM3-1.7B (free) as a compact 1.7B-parameter text model with a 262K-token context window, released Nov 6, 2026, and post-trained for agentic work, tool calling, and structured outputs. The listing notes it reasons before producing an answer, and marks tool calling as supported for the text-only model. The catalog entry identifies the exact variant as liquid/lfm-3-1.7b:free and lists a broad set of supported API parameters, including reasoning controls, structured output and response format, tool choice, seed, and sampling options such as temperature and top_p. It also flags that prompts and outputs may be retained and used, a detail worth checking before routing production traffic through the free tier.

LiquidAI’s LFM3-1.7B is listed as a compact 1.7B-parameter text model with a 256K-token context, 8.2K maximum output, and post-training for agents, tool calling, structured outputs, and reasoning before answering. The catalog attributes the model to Liquid and marks both the standard and free identifiers as new in November 2026. The same listing identifies the exact free variant, liquid/lfm-3-1.7b:free, with free input and output pricing alongside text-to-text operation and tool-calling, reasoning, and structured-output support. It also warns that prompts and outputs may be retained and used, a relevant operational detail for developers testing the model.

An independent pricing and benchmark dashboard tracks 379 models and flags LiquidAI: LFM3-1.7B as new for the week of 2026-11-06, with a 262K context window. It records the free variant liquid/lfm-3-1.7b:free alongside Google: Gemini Omni 1.2 Flash, released 2026-11-05. The dashboard reports per-model price points derived from tracked data, including A$0.12 in and A$0.29 out for the standard LFM3-1.7B listing, while the free variant is shown with free input and output. It notes that benchmark scores are best published mapped runs rather than universal task results, and that reasoning support is a capability flag rather than a reasoning benchmark.

OpenRouter's Liquid provider page lists the exact free variant LiquidAI: LFM3-1.7B (free), slug liquid/lfm-3-1.7b:free, released Nov 6, 2026 with a 262K context and $0/M input and output tokens. It is described as a compact 1.7B-parameter text model post-trained for agents, tool calling, and structured outputs. The same page positions the free variant alongside the paid sibling LFM3-1.7B at $0.08/M input and $0.20/M output, and shows roughly 3.05B tokens processed for the free version versus 6.33B for the paid one. It also lists the d1.1 structured decision model, a System One endpoint returning choices, scores, and yes/no answers with probabilities, dated Oct 27, 2026.

LiquidAI: LFM3-1.7B (free) is a compact 1.7B-parameter text model with a 256K-token context, released Nov 6, 2026, post-trained for agents, tool calling, and structured outputs. It reasons before it answers, and OpenRouter lists it as a free variant served by a single provider with no routing decisions involved. Observed serving data on the OpenRouter model page shows 167 tok/s throughput and 0.29s P50 latency, with 92.98% uptime and 92.05% three-day availability. The page also reports a 5.19% tool call error rate and a notably high 73.20% structured output error rate, plus a disclosure that prompts and outputs may be retained and used to train Liquid models.