KVARK Sparkle 1.0 (27B) FP8

FP8 weights of KVARK Sparkle 1.0 for vLLM, with the multi-token prediction (MTP) head. Every number on this card was measured with these weights. The files take 29 GB, and one H200 serves them with about 128k context.

Results

Both models ran with FP8 weights, thinking on, a 64k token cap and the same prompts. Sampling was temperature 1.0, top_p 0.95 and top_k 20. Differences are paired over items, with 95% bootstrap intervals.

Benchmark	Accuracy (%)			Median reasoning tokens			64k cap hits
Benchmark	Sparkle	Qwen	Difference	Sparkle	Qwen	Cut	64k cap hits
GPQA Diamond198 items × 5 seeds	89.3	89.4	−0.1[−1.9, +1.8]	3,122	7,854	−60%	9 / 34
AIME2025 and 2026 · 60 problems × 5 seeds	93.3	94.0	−0.7[−3.3, +2.0]	5,502	14,318	−62%	12 / 12
LiveCodeBenchv6 · 50 recent problems × 3 seeds	82.7	77.3	+5.3[0.0, +11.3]	6,664	19,950	−67%	15 / 27
IFEval541 prompts, strict	91.3	92.4	−1.1[−2.4, +0.2]	470	714	−34%	0 / 0
MMLU-Pro980 items, 70 per category	83.3	84.0	−0.7[−2.2, +0.8]	429	727	−41%	0 / 5

Speed

One user, 1 × H200	Sparkle, MTP	Qwen, MTP	Qwen, no MTP
Median time per answer	13.7 s	28.1 s	49.1 s
MTP mean acceptance length	2.53	2.50	off

Measured with vLLM 0.30 on 20 fixed questions (10 GPQA Diamond, 10 LiveCodeBench), one request at a time, thinking on.

Serve with vLLM

vllm serve Egzakta/Kvark-Sparkle-1.0-27B-FP8 \
  --max-model-len 131072 --kv-cache-dtype fp8 \
  --reasoning-parser qwen3 \
  --enable-auto-tool-choice --tool-call-parser qwen3_coder \
  --speculative-config '{"method":"mtp","num_speculative_tokens":3}'

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="none")
response = client.chat.completions.create(
    model="Egzakta/Kvark-Sparkle-1.0-27B-FP8",
    messages=[{"role": "user", "content": "How many primes are there between 100 and 150?"}],
    max_tokens=65536,
)
message = response.choices[0].message
print(getattr(message, "reasoning", None))
print(message.content)
print(response.usage.completion_tokens, "tokens")

Thinking is on by default. To turn it off for a request, pass extra_body={"chat_template_kwargs": {"enable_thinking": False}}. Sampling defaults are temperature 1.0, top_p 0.95 and top_k 20.

Memory

Context	Memory per sequence, FP8 KV cache
32,768	1 GiB
65,536	2 GiB
131,072	4 GiB
Linear-attention state, any context	under 0.2 GB

Training, data and the BF16 weights are on the main model card. License: Apache 2.0, commercial use permitted, no revenue cap.

Egzakta/Kvark-Sparkle-1.0-27B-FP8