iFlytek open-sources Spark-X2.6-4B: a 4B on-device model with 1M context, beating Qwen3.5-9B
AdinaYakup · x · 2026-10-22
SparkLLM released Spark-X2.6-4B, an Apache 2.0 on-device model supporting up to 1M-token context, three thinking modes, and local deployment via GGUF.
The accompanying benchmark chart shows it leading Qwen3.5-9B and Gemma4-12B/27B across agent, tool-use, and long-context benchmarks including BFCL-V4 (71.3), MCP-Atlas (62.4), BrowseComp (53.1), SWE Pro (54.3), and HLE w/ tools (38.9).
Related event: iFlytek Open-Sources Spark-X2.6-4B Edge Model with 1M Context(4 posts)→
More from Infra
- Claude Models on Google Cloud (Anthropic managed) is now generally available — claudeai · 2026-10-23
- Scaling GLM-5.3-Flash training to 1M token context, 64x over Megatron baseline — AndrewLBeam · 2026-10-23
- Perplexity's Portable Computer lands on AMD Ryzen AI Max, running local 27B model — AravSrinivas · 2026-10-23
- Liquid AI Launches RIPL: Ultra-Low-Latency Inference Engine With Sub-50ms End-to-End Responses — JosephJacks_ · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Jensen Huang lays out NVIDIA's full Vera Rubin agentic system: NVL72, MGX, CMX and RTX — rohanpaul_ai · 2026-10-23