whatmodel.sh / Spark-X2.6 4B

Spark-X2.6 4B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Spark-X2.6 4B is an open-weights language model from iFLYTEK (SparkLLM), released October 2026. Its published size is 4.112B parameters and a 1M-token context window. Licence: apache-2.0.

Catalogue identity notes

Post-trained Spark-X2.6 checkpoint, distinct from X2.5 and Base. Published evaluations use thinking high, temperature 1.0, top_p 0.95, top_k -1. Quantized download quality and speed were not measured in this update.

Model information sources

Release date ↗ · Weights licence ↗ · Parameter count ↗

Hugging Face safetensors metadata: 4,112,079,360 parameters.

67.2±1.0#111 of 732 LLMsfrom 17 benchmark results across 1 source

Score within the LMArena (text) comparison set only.

$ how it compares

The LLMs ranked either side of Spark-X2.6 4B, out of 886 with a score, then more from iFLYTEK (SparkLLM).

#modelscorewhereRAMreleased
124MiMo-V2-Omni Xiaomi67±2comparison only2026-03
125Nemotron 3.5 Super VL 120B A12B NVIDIA67±1local~254GB2026-10
126Spark-X2.6 4B iFLYTEK (SparkLLM)67±1local~5.2GB2026-10
127DeepSeek V3.2 DeepSeek67±1api $0.29/M +local~110GB2025-12
128Step 3.5 Flash (196B-A11B) StepFun67±1api $0.15/M +local~118GB2026-02
more from iFLYTEK (SparkLLM)
169Spark-X2.5 4B iFLYTEK (SparkLLM)65±1local~3.5GB2026-08
260Spark-X2.6 1.7B iFLYTEK (SparkLLM)59±1local~3.4GB2026-10
347Spark-X2.5 1.7B iFLYTEK (SparkLLM)55±1local~1.9GB2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▄▅▅▄▄▂▁ #94 of 313 · 31% of its score · 7 tests

Beats Grok 4.20 (0309 reasoning), GPT-5.1, o3-pro here, though they rank higher overall. Trails Hy3 preview (295B-A21B), MiniMax-M2.7, Step 3.5 Flash (196B-A11B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab54.3% ✎▂·▁▁▁▁▂▃█▇▂▁ beats 32/58very hard · avg 9.8%▇▇▇▇ 10%sharp · at its level
BrowseComp lab53.1% ✎▂·▃▁▂▂▆▅█▂▂▄ beats 22/46hard · avg 21%▇▇ 4.1%at its level
SWE-bench Verified lab62.8% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 64/138hard · avg 19%▇▇ 3.9%sharp · at its level
SWE-bench Multilingual lab64.7% ✎▁··▁▃▁▅██▂▂▁ beats 28/65hard · avg 25%▇ 3.5%at its level
Humanity's Last Exam with tools lab36.9% ✎▁·▁▂▂▁▄▄▇█▄▅ beats 18/62very hard · avg 11%▇ 3.4%at its level
BFCL V4 (December 2025 archive) lab71.8% ✎▁▁▁▂▆▆▅▆█▂▄▃ beats 108/114medium · avg 40%▇ 3.0%at its level
τ²-bench aggregate lab86.9% ✎▁▁▁·▁▁▃▃▃▄█▁ beats 34/39medium · avg 54%▇ 2.6%typical
coding▁▁▁▃▅▇█▇▅▄▃▁ #101 of 469 · 22% of its score · 4 tests

Beats Grok 4.20 (0309 reasoning), Grok 4.3, Muse Glimmer 30B here, though they rank higher overall. Trails Llama 3.1 Nemotron Ultra 253B v1, Qwen3 235B-A22B Thinking 2507, o4-mini here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab54.3% ✎▂·▁▁▁▁▂▃█▇▂▁ beats 32/58very hard · avg 9.8%▇▇▇▇ 10%sharp · at its level
SciCode (level unspecified) lab36.8% ✎▁····▃▃▄▅█▆▂ beats 22/40very hard · avg 14%▇▇ 4.2%at its level
SWE-bench Verified lab62.8% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 64/138hard · avg 19%▇▇ 3.9%sharp · at its level
SWE-bench Multilingual lab64.7% ✎▁··▁▃▁▅██▂▂▁ beats 28/65hard · avg 25%▇ 3.5%at its level
chat & writing▁▁▁▃▆▄▆█▆▇▇▃ #100 of 402 * · 2.1% of its score · 1 test

Beats LongCat 2.0 (1.6T-A48B), Inkling Small (266B-A12B), GPT-5.4 nano here, though they rank higher overall. Trails Gemma 4 12B, Kimi K2 Instruct (0905), Gemma 4 26B-A4B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
IFEval (aggregation unspecified) lab93.6% ✎▁▂▄▂▂▄▆██▅▂▁ beats 135/145easy · avg 83%▇ 2.1%maxed out
math▁▁▃▅█▆▄█▄▂▁▁ #95 of 431 · 18% of its score · 3 tests

Beats Claude Sonnet 4.6, Claude Opus 4.5, MiniMax-M3 here, though they rank higher overall. Trails Gemini 2.5 Flash, Granite 4.2 30B, Qwen3-Next 80B-A3B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HMMT February 2026 lab81.3% ✎▃▃▂▅·████▅▃▅ beats 22/35hard · avg 32%▇▇▇ 6.6%sharp
IMO-AnswerBench lab75.2% ✎▁▁·▁▁▁▄▆█▄▂▃ beats 15/44hard · avg 32%▇▇ 6.2%typical
AIME 2026 lab91.2% ✎▁▁▁▂▂▂▆█▆▂·▁ beats 27/52medium · avg 53%▇▇ 5.4%sharp
web search▅·▅▅█·▃▃▃·▅▃ #1 of 15 * · 7.5% of its score · 2 tests
testresultvs the others in its rundifficultycountswhy
BrowseComp lab53.1% ✎▂·▃▁▂▂▆▅█▂▂▄ beats 22/46hard · avg 21%▇▇ 4.1%at its level
Humanity's Last Exam with tools lab36.9% ✎▁·▁▂▂▁▄▄▇█▄▅ beats 18/62very hard · avg 11%▇ 3.4%at its level
other tests5 not in a task · 44% of its score
testresultvs the others in its rundifficultycountswhy
GPQA Diamond lab77.1% ✎▁▁▁▁▂▄▅▅███▄ beats 190/322hard · avg 38%▇▇▇▇▇▇ 14%sharp · at its level
GPQA (subset unspecified) lab69.9% ✎▂▁▂▂▄▃▆▃▅█▃▃ beats 77/106hard · avg 37%▇▇▇▇ 11%sharp · at its level
Humanity's Last Exam lab17.9% ✎▁▁▁▄▆▆▇██▆▁▃ beats 45/111very hard · avg 3.8%▇▇▇ 8.3%typical
LongBench v2 lab53.3% ✎▂·▁·▂▃▁▄▄▃█▄ beats 19/33very hard · avg 32%▇▇▇ 7.9%at its level
IFBench (aggregation unspecified) lab75.8% ✎▂▆▂▄▄▄▂▄▄▆█▇ beats 68/86medium · avg 49%▇ 3.0%typical

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run Spark-X2.6 4B · needs ~3.5434 GB · from huggingface.co/XHToken
    chat in the terminal
    llama cli -hf XHToken/Spark-X2.6-4B-GGUF
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf XHToken/Spark-X2.6-4B-GGUF

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Spark-X2.6 4B · needs ~3.5434 GB · from huggingface.co/XHToken

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/XHToken/Spark-X2.6-4B-GGUF -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Spark-X2.6 4B · needs ~3.5434 GB · from huggingface.co/XHToken
    chat in the terminal
    ollama run hf.co/XHToken/Spark-X2.6-4B-GGUF

Model repository and documentation: huggingface.co/XHToken ↗

$ how the score was worked out · Spark-X2.6 model cards (think… 100%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Spark-X2.6 model cards (thinking high) · maker-reported17Apache-2.0
prior · pull toward the middle
weighted average= 67.2

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json