Up to 3.2x Faster Inference with LFM2.5-DSpark

Up to 3.2x Faster Inference with LFM2.5-DSpark Up to 3.2x Faster Inference with LFM2.5-DSpark Team Article Published August 20, 2026 Upvote 56 xx tugot17 LiquidAI Leonie Monigatti iamleonie LiquidAI Fernando Fernandes Neto fernandofernandes LiquidAI Tarek Dakhran tdakhran LiquidAI nathan ranchin nathanrchn LiquidAI Today, we release DSpark draft model checkpoints for three models from our LFM2.5 family: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B. These add a speculative decoding path that trades a minimal memory increase for a large decoding speedup without changing output quality: Faster inference : up to 3.18 throughput improvement on a GPU and up to 2.87x on-device.

This HardwareLaunch is relevant to the technology intelligence record because it involves Meta, OpenAI, Hugging Face, llama. The source article should remain the factual reference for follow-up coverage.

• Up to 3.2x Faster Inference with LFM2.5-DSpark Team Article Published August 20, 2026 Upvote 56 xx tugot17 LiquidAI Leonie Monigatti iamleonie LiquidAI Fernando Fernandes Neto fernandofernandes LiquidAI Tarek Dakhran tdakhran LiquidAI nathan ranchin nathanrchn LiquidAI Today, we release DSpark draft model checkpoints for three models from our LFM2.5 family: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B.
• These add a speculative decoding path that trades a minimal memory increase for a large decoding speedup without changing output quality: Faster inference : up to 3.18 throughput improvement on a GPU and up to 2.87x on-device.
• Toward on-device agentic inference : cuts function-calling latency by 57% on average for LFM2.5-2.6B Day-one support for llama.cpp and SGLang : LFM-compatible DSpark integration is open-sourced upstream How does DSpark work The decode phase in LLM inference is traditionally memory-bound.
• Most latency comes from streaming weights from DRAM into SRAM, not from intense computation.
• Speculative decoding addresses this by using a lightweight draft model to produce candidate tokens, then having the target model verify them all in a single forward pass, sharing the cost of loading the weights across all tokens we verify.
• Over the years, multiple approaches of speculation have been proposed, with the most prominent being EAGLE-3 , DFlash , and, most recently, DSpark , which combines three components: DFlash-style parallel backbone conditioned on the target model’s context features, producing hidden states for all draft tokens in a single forward pass.