iFlytek open-sources Spark-X2.6-1.7B with 1M context and leading on-device benchmarks
AdinaYakup · x · 2026-10-27
iFlytek's SparkLLM followed up its 4B model with the smaller Spark-X2.6-1.7B, targeting on-device use:
- Three thinking modes (none/low/high), up to 1M context, Apache 2.0 license, with GGUF/int4/int8/FP8 quantizations.
- Official benchmark chart shows the 1.7B averaging 49.6 across Agent/Code/Maths categories, ahead of Qwen3.5 2B, Gemma4 E2B, MiniCPM5 2B and LFM2.5-2.6B; the 4B variant averages 63.3.
- The HF collection lists 4B and 1.7B tiers with Base and multiple quantized variants; the 4B GGUF has passed 100k downloads.
Related event: iFlytek Open-Sources Spark-X2.6-1.7B with 1M Context(2 posts)→
More from Models
- Statistically lossless Qwen3.8 27B GGUF quants drop, UD Q8 matches BF16 at 99.5% — udmrzn · 2026-10-27
- Mistral Unveils 'Le Chonk' Model, Throws Launch Parties in SF, NYC and London — b_roziere · 2026-10-27
- EmbeddingGemma2 powers a 20ms real-time endpoint in this dev's Jev project — bodonoghue85 · 2026-10-27
- Qoder offers DeepSeek-V4.1-Flash free for all accounts through Oct 31 — Div_pradeep · 2026-10-27
- Upgrade Codex Dreaming V3 memory with two config lines for smarter, cheaper models — daniel_mac8 · 2026-10-27
- ProgramBench goes from all-zero to Anthropic leading at 18% in months — jyangballin · 2026-10-27