Spark-X2.6
Spark-X2.6 is a pair of compact 4B and 1.7B open-weight language models built for on-device agents, sharing Spark-X2.5's architecture but with new post-training that improves tool use, multi-step planning, coding, long context and multilingual tasks. Apache-2.0.
Updated
✨ Try AgentScout's free AI image generator →Good for: writing code, training and fine-tuning agents
Use cases
Boost Spark-X2.6 with our $29 Fast-Track + 7-Day Launch Boost, pin it to the top of models & local runtimes for 30 days + permanent /best/ & /compare/ placement + dual X spotlight (@clampdown19 & @AgentScoutHQ) for $49, or book the exclusive $99/wk sitewide & newsletter slot.
About Spark-X2.6
Spark-X2.6 is listed in our models & local runtimes category and is aimed at builders. Spark-X2.6 is a pair of compact 4B and 1.7B open-weight language models built for on-device agents, sharing Spark-X2.5's architecture but with new post-training that improves tool use, multi-step planning, coding, long context and multilingual tasks. Apache-2.0. Pricing, as the maker states it: Open weights. Maturity: Active. We last checked the official link on 2026-10-29.
Spark-X2.6 at a glance
- Category: Models & local runtimes
- Built for: Developers and teams building with AI agents
- Pricing: Open weights (as the maker states it)
- GitHub: XHToken/Spark-X2.6 · 864 stars (as of Oct 29, 2026)
- Official site: github.com/XHToken/Spark-X2.6
- Listed on AgentScout: Oct 29, 2026
Spark-X2.6 FAQ
What is Spark-X2.6?
Spark-X2.6 is a pair of compact 4B and 1.7B open-weight language models built for on-device agents, sharing Spark-X2.5's architecture but with new post-training that improves tool use, multi-step planning, coding, long context and multilingual tasks. Apache-2.0.
How much does Spark-X2.6 cost?
Spark-X2.6's pricing, as the maker states it: Open weights. Open-source tools are free to use, though you may still pay for model API usage or a hosted plan. Check the official site for current plans.
Where can I find Spark-X2.6's source code?
On GitHub at XHToken/Spark-X2.6, which had 864 stars as of Oct 29, 2026.
What are alternatives to Spark-X2.6?
Similar tools include MiroThinker, GLM-5, Meta Muse Wonder, Mercury 3, Ollama and StartLux Decision.
Facts from our link-checked listing, the GitHub repository and the official site. See something wrong? Suggest an edit.
Spark-X2.6 ratings and reviews
Be the first to rate Spark-X2.6: tap 1 to 5 stars below. A written review is optional.
Spark-X2.6 alternatives
Other models & local runtimes tools in our directory. See all models & local runtimes tools →
MiroThinker
Deep research agent and open models tuned for complex research and forecasting tasks.
GLM-5
Z.ai's open-weights GLM-5 model family built for coding and long-horizon agentic tasks, also offered through the Z.ai API.
Meta Muse Wonder
Meta's new models for personal and professional agent tasks, available via Meta Model API and Muse Code.
Mercury 3
Mercury 3 is Inception's diffusion language model, which writes many tokens in parallel rather than one at a time, reaching over 1,000 tokens per second on NVIDIA GPUs with a 1M-token context, tool calling and structured outputs. It has an OpenAI-compatible API and is also offered on OpenRouter, Baseten and Vercel, which makes it a fit for voice bots, support agents and browser agents where response speed matters. Pricing is per million input and output tokens.
Ollama
Runs open models such as Qwen, Gemma, DeepSeek and gpt-oss locally or in Ollama's cloud, and connects them to agents like Claude Code, OpenClaw and Codex.
StartLux Decision
Open decision models (0.8B to 27B) that return a calibrated probability for every option in a single pass, with an inference engine and HTTP server.
Related lists
🏆 Ranked in: LocalAI alternatives
🔁 Listed in: LocalAI alternatives
🧭 Task guides: Planning and task tracking tools for coding agents
💲 Estimate API cost: compare per-request and monthly costs for GPT, Claude, Gemini and Grok models.
Badge for makers
Built Spark-X2.6? Add this badge to your README or site. It links back to this page. About badges
