Skip to content
Other

xspark-X2.6-math

4B-математик без reasoning-режима

1 min readupdated

Экспериментальный математический чекпойнт на базе Spark-X2.6-4B от arcticoneai. 4B параметров, rank-16 LoRA уже слит в BF16-веса, отдельный адаптер не нужен. Ориентирован на решение задач и формализацию в Lean 4.

Ключевые особенности

  • AIME: 50% у базовой Spark-X2.6-4B и 90% у math-версии — оба замера без включённого reasoning и agent harness.
  • 90% получены на внутренней выборке из 50 вопросов, это не аудированный лидерборд.
  • В предоставленном Lean 4-фрагменте сгенерированы доказательства для пары неизоморфных групп с изоморфными tracial group factors и для бесконечного семейства таких групп.
  • Дообучение: rank-16 LoRA, слитая в BF16, плюс дистилляция reasoning от учителя с восемью специализированными промптами.
  • Карточка называет чекпойнт экспериментальным: нужны полный Lean-проект и аудит зависимостей.

Пример

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = 'arcticoneai/xspark-X2.6-math'
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id, torch_dtype='auto', device_map='auto', trust_remote_code=True
)

Используйте формат thinking/reasoning, ожидаемый базовой Spark-X2.6.

ПодробнееAIME: 50% → 90% без reasoning, но 90% — внутренняя выборка из 50 вопросов.
Lean 4: две теоремы о группах с изоморфными tracial factors; фрагмент сокращён.
Для независимой проверки нужны полный Lean-проект, аудит аксиом и скрипт оценки.
Модель: 4B параметров, LoRA rank-16 слита в BF16.
Discuss Telegram 3 223

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 7.7 GB · ready files8.9 GB
  • FP8 / INT8recommendedweights 3.8 GB4.7 GB
  • 4-bitweights 2.2 GB2.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 3080

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # arcticoneai/xspark-X2.6-math
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download arcticoneai/xspark-X2.6-math --local-dir ~/models/xspark-X2.6-math
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to xspark-X2.6-math
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Spark-X2.6-4B-Basefamily root
  2. Spark-X2.6-4B
  3. xspark-X2.6-math

New posts about versions of Spark-X2.6-4B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 7All on Hugging Face
Fine-tunes 5All on Hugging Face

About the model

arcticoneai/xspark-X2.6-math
Author
arcticoneai
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

2233

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Other

ice40GPT

ice40GPT