// head_to_head

GPT-6 Astra vs Skyfall 36B V2

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

Skyfall 36B V2 is the cheaper of the two; neither can be ranked on quality here.

Skyfall 36B V2 does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.

openai

GPT-6 Astra

Blended / 1M
$20.00
Context
1.1M
Released
Sep 4, 2026
Overall score
82.2
reasoningtool callingfile inputimage inputprompt caching

thedrummer

Skyfall 36B V2

Blended / 1M
$0.613
Context
33K
Released
Mar 10, 2025
Overall score
Not evaluated
prompt caching

Specs and pricing

MetricGPT-6 AstraSkyfall 36B V2
LiveBench overall

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

82.2
Cost per point

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

$0.3942
Blended price / 1M

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

$20.00$0.613win
Input price / 1M$10.00$0.550win
Output price / 1M$50.00$0.800win
Cached input / 1M

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

$1.00$0.250win
Context window1.1Mwin33K
Max output tokens128Kwin29K

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — GPT-6 Astra on top, Skyfall 36B V2 below, both out of 100.

Agentic coding
57.3
Coding
80.4
Reasoning
92.7
Mathematics
96.8
Data analysis
83.0
Language
89.4
Instruction following
75.6

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.

WorkloadGPT-6 AstraSkyfall 36B V2
Support chatbot

1.2K in / 400 out × 200K requests

$5,752/mo$174.40/mo
RAG assistant

8K in / 600 out × 100K requests

$7,400/mo$368.00/mo
Coding agent

40K in / 4K out × 20K requests

$6,960/mo$336.00/mo
Document extraction

20K in / 1.5K out × 50K requests

$13,300/mo$595.00/mo
Bulk classification

500 in / 20 out × 5M requests

$25,500/mo$1,305/mo
Run these two through the cost calculator

Which should you pick?

You need to fit large documents in one call

GPT-6 Astra

Wider context window — 1.1M against 33K.

You are cost-constrained

Skyfall 36B V2

Cheaper on blended list price at $0.613 per million tokens.

GPT-6 Astra vs Skyfall 36B V2 FAQ

Which is better, GPT-6 Astra or Skyfall 36B V2?

Skyfall 36B V2 is the cheaper of the two; neither can be ranked on quality here. Skyfall 36B V2 does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.

Is GPT-6 Astra cheaper than Skyfall 36B V2?

Skyfall 36B V2 is cheaper. On a 3:1 input:output blend, GPT-6 Astra lists at $20.00 per million tokens and Skyfall 36B V2 at $0.613 — Skyfall 36B V2 is 33× cheaper. Input and output are priced separately — GPT-6 Astra charges $10.00 in and $50.00 out, Skyfall 36B V2 charges $0.550 and $0.800 — so the model that looks cheaper flips depending on how output-heavy your workload is.

Does GPT-6 Astra or Skyfall 36B V2 have a bigger context window?

GPT-6 Astra has the larger context window: 1.1M for GPT-6 Astra against 33K for Skyfall 36B V2. 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 GPT-6 Astra and Skyfall 36B V2 support prompt caching?

Both publish a cached-input rate: $1.00 per million for GPT-6 Astra and $0.250 for Skyfall 36B V2, against full input rates of $10.00 and $0.550. 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.
  • ScoresLiveBench release 2026-06-25, 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.