vakati-tech.ai

// head_to_head

Qwen3.8 27B vs SpaceXSI: Grok 4.6

Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.

OpenRouter + LiveBenchAll comparisonsFull leaderboard

SpaceXSI: Grok 4.6 scores higher, Qwen3.8 27B costs less — it depends on your workload.

SpaceXSI: Grok 4.6 is ahead by 9.5 points overall, and Qwen3.8 27B lists 5.6× cheaper per blended million tokens. Whether 9.5 points is worth that depends on how much a wrong answer costs you. Qwen3.8 27B also leads on measured cost per point of capability, at $0.0106 per point.

qwen

Qwen3.8 27B

Blended / 1M
$0.535
Context
1.0M
Released
Aug 14, 2026
Overall score
54.0
reasoningtool callingimage inputvideo inputprompt caching

x-ai

SpaceXSI: Grok 4.6

Blended / 1M
$3.00
Context
500K
Released
Aug 12, 2026
Overall score
63.5
reasoningtool callingimage inputfile inputprompt caching

Specs and pricing

MetricQwen3.8 27BSpaceXSI: Grok 4.6
LiveBench overall

Weighted mean of the seven LiveBench category scores, 0–100. Higher is better.

54.063.5win
Cost per point

Measured benchmark spend divided by overall score — dollars per point of capability.

$0.0106win$0.0192
Blended price / 1M

3:1 input:output mix, the usual shape of production traffic.

$0.535win$3.00
Input price / 1M$0.060win$2.00
Output price / 1M$1.96win$6.00
Cached input / 1M

Price of an input token served from the prompt cache, where the provider publishes one.

$0.059win$0.500
Context window1.0Mwin500K
Max output tokens236K450Kwin

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Qwen3.8 27B on top, SpaceXSI: Grok 4.6 below, both out of 100.

Long-horizon agentic
15.9
18.7
Coding
69.3
76.7
Reasoning
80.0
91.2
Browser
24.2
49.4
Data analysis
76.6
75.3
Research
61.3
88.8
Documents & design
72.4
76.3

What each one costs to run

Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.

WorkloadQwen3.8 27BSpaceXSI: Grok 4.6
Support chatbot

1.2K in / 400 out × 200K requests

$171.13/mo$852.00/mo
RAG assistant

8K in / 600 out × 100K requests

$165.20/mo$1,360/mo
Coding agent

40K in / 4K out × 20K requests

$204.24/mo$1,240/mo
Document extraction

20K in / 1.5K out × 50K requests

$206.95/mo$2,375/mo
Bulk classification

500 in / 20 out × 5M requests

$345.50/mo$4,850/mo
Run these two through the cost calculator

Which should you pick?

You are running this at volume

Qwen3.8 27B

Lowest measured cost per point of capability at $0.0106 per point — the gap compounds with every request.

Quality matters more than the bill

SpaceXSI: Grok 4.6

Highest overall LiveBench score of the two at 63.5.

The workload is coding or agentic work

SpaceXSI: Grok 4.6

Leads on long-horizon agentic — 18.7 against 15.9.

You need to fit large documents in one call

Qwen3.8 27B

Wider context window — 1.0M against 500K.

Qwen3.8 27B vs SpaceXSI: Grok 4.6 FAQ

Which is better, Qwen3.8 27B or SpaceXSI: Grok 4.6?

SpaceXSI: Grok 4.6 scores higher, Qwen3.8 27B costs less — it depends on your workload. SpaceXSI: Grok 4.6 is ahead by 9.5 points overall, and Qwen3.8 27B lists 5.6× cheaper per blended million tokens. Whether 9.5 points is worth that depends on how much a wrong answer costs you. Qwen3.8 27B also leads on measured cost per point of capability, at $0.0106 per point.

Is Qwen3.8 27B cheaper than SpaceXSI: Grok 4.6?

Qwen3.8 27B is cheaper. On a 3:1 input:output blend, Qwen3.8 27B lists at $0.535 per million tokens and SpaceXSI: Grok 4.6 at $3.00 — Qwen3.8 27B is 5.6× cheaper. Input and output are priced separately — Qwen3.8 27B charges $0.060 in and $1.96 out, SpaceXSI: Grok 4.6 charges $2.00 and $6.00 — so the model that looks cheaper flips depending on how output-heavy your workload is.

Qwen3.8 27B vs SpaceXSI: Grok 4.6: which scores higher on benchmarks?

Qwen3.8 27B scores 54.0 and SpaceXSI: Grok 4.6 scores 63.5 overall on LiveBench, the weighted mean of its seven categories. That is a 9.5-point lead for SpaceXSI: Grok 4.6. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.

Which gives better value for money, Qwen3.8 27B or SpaceXSI: Grok 4.6?

Qwen3.8 27B. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. Qwen3.8 27B works out at $0.0106 per point and SpaceXSI: Grok 4.6 at $0.0192.

Does Qwen3.8 27B or SpaceXSI: Grok 4.6 have a bigger context window?

Qwen3.8 27B has the larger context window: 1.0M for Qwen3.8 27B against 500K for SpaceXSI: Grok 4.6. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.

Do Qwen3.8 27B and SpaceXSI: Grok 4.6 support prompt caching?

Both publish a cached-input rate: $0.059 per million for Qwen3.8 27B and $0.500 for SpaceXSI: Grok 4.6, against full input rates of $0.060 and $2.00. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.

Related comparisons

How these numbers are produced

  • Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
  • Scores — LiveBench release 2026-11-06, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
  • Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
  • “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.

Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.