GPT-6 Astra · Math

GPT-6 Astra for math: our test runs

GPT-6 Astra ran our 5 math prompts in llmwise, through the same pipeline your messages take. Here's every reply as it came, whether it passed, what it cost and what it counts as on Pro, next to the other models.

Start free See pricing

Based on 5 of our test runs on September 27, 2026, through OpenRouter with the app's own prompt and settings. Updated October 15, 2026.

Short answer

In our test runs on September 27, 2026, GPT-6 Astra passed 5 of 5 of our math prompts (both hard ones): tied 1st of 20 models. Each reply counted as 1 message on Pro, where it has 31 a month, and cost $0.0085 to run on average.

Prompt by prompt

Each of the 5 math prompts, and how GPT-6 Astra's reply did.

GPT-6 Astra on each math prompt
Prompt	Result	Cost	Time
A discount, then sales taxEveryday	Passed: Final answer 66: right.	$0.0063	1.9 s
Pens at 3 for $4Everyday	Passed: Final answer Buy three groups of 3 and one single pen for 13.50.: right.	$0.0082	2.4 s
Compound interest over three yearsEveryday	Passed: Final answer 315.25: right.	$0.0072	2.7 s
Four-digit numbers whose digits sum to 9Hard	Passed: Final answer 165: right.	$0.0122	3.8 s
The highest of three dice is a 5Hard	Passed: Final answer 61/216: right.	$0.0088	2.6 s

Every math reply from GPT-6 Astra

A discount, then sales tax
An everyday prompt: “A discount, then sales tax”, in full.
Passed: Final answer 66: right.
```
After the 25% discount: $80 × 0.75 = $60.  
Adding 10% sales tax: $60 × 1.10 = $66.

Final answer: $66
```
400 tokens in, 46 out · 1.9 s · $0.0063 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI · September 27, 2026
Pens at 3 for $4
An everyday prompt: “Pens at 3 for $4”, in full.
Passed: Final answer Buy three groups of 3 and one single pen for 13.50.: right.
```
A group of 3 costs $4, compared with $4.50 for 3 singles, so buy as many groups as possible.

10 pens = 3 groups of 3 + 1 single.
Cost = 3 × $4 + $1.50 = $13.50.

Final answer: Buy three groups of 3 and one single pen for $13.50.
```
402 tokens in, 83 out · 2.4 s · $0.0082 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI · September 27, 2026
Compound interest over three years
An everyday prompt: “Compound interest over three years”, in full.
Passed: Final answer 315.25: right.
```
After 3 years, the balance is:
\[
2000(1.05)^3 = \$2,315.25
\]

Interest earned:
\[
\$2,315.25 - \$2,000 = \$315.25
\]

Final answer: $315.25
```
402 tokens in, 64 out · 2.7 s · $0.0072 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI · September 27, 2026

Four-digit numbers whose digits sum to 9

A hard prompt: “Four-digit numbers whose digits sum to 9”, in full.

Passed: Final answer 165: right.

Let the digits be \(a,b,c,d\), where \(a\ge1\) and the others are nonnegative. We need
\[
a+b+c+d=9.
\]
Set \(x=a-1\). Then \(x+b+c+d=8\). By stars and bars, the number of nonnegative solutions is
\[
\binom{8+4-1}{4-1}=\binom{11}{3}=165.
\]
All solutions give valid digits, since \(a\le9\) and \(b,c,d\le8\).

Final answer: 165

378 tokens in, 169 out (35 of them reasoning) · 3.8 s · $0.0122 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI · September 27, 2026

The highest of three dice is a 5

A hard prompt: “The highest of three dice is a 5”, in full.

Passed: Final answer 61/216: right.

There are \(6^3=216\) equally likely outcomes. For the highest number to be exactly 5, all dice must show at most 5, but not all at most 4.

Thus,
\[
P(\text{highest}=5)=\frac{5^3-4^3}{6^3}
=\frac{125-64}{216}
=\frac{61}{216}.
\]

Final answer: 61/216

393 tokens in, 97 out · 2.6 s · $0.0088 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI · September 27, 2026

How it compares on math

GPT-6 Astra beside the job's picks among every model. 19 models share the hard-problems pick, GPT-6 Astra among them.

GPT-6 Astra beside other models on math
Model	Passed	On Pro	Cost per reply
GLM 5.3Our pick: hard problems (shared), best value	5 of 5	250 a month on Pro	$0.0004
Claude Haiku 5.5Our pick: hard problems (shared), everyday	5 of 5	60 a day on Pro	$0.0001
GPT-6 AstraOur pick: hard problems (shared)	5 of 5	31 a month on Pro	$0.0085

On Pro: Pro's count on each model. Cost: what OpenRouter charged us per reply, on average.

Every limit is published. Paid plans also have a monthly fair-use limit on AI cost: Plus $3.75, Pro $7.50, Max $20, Ultra $42, Studio $85. Using every message on your plan at typical sizes stays under it; very large messages and heavy research use it faster. Every limit, explained.

GPT-6 Astra next to each model on math

GPT-6 Astra against each model on the same 5 math prompts. One run each, through OpenRouter: a wait depends on the provider and the load that day, so a lead under 10% counts as close.

Against Claude Fable 5.1: GPT-6 Astra answered 1.8× sooner (2.6 s to 4.6 s), cost 1.4× less, and passed 5 of 5 to its 5.
Against Claude Opus 5.5: GPT-6 Astra answered 1.5× sooner (2.6 s to 3.9 s), cost 1.4× more, and passed 5 of 5 to its 5.
Against Claude Sonnet 5.5: GPT-6 Astra answered 1.6× later (2.6 s to 1.7 s), cost 3.5× more, and passed 5 of 5 to its 5.
Against Claude Sonnet 5: GPT-6 Astra answered 1.5× sooner (2.6 s to 3.9 s), cost 2.6× more, and passed 5 of 5 to its 5.
Against Claude Haiku 5.5: GPT-6 Astra answered 1.5× later (2.6 s to 1.7 s), cost 58.1× more, and passed 5 of 5 to its 5.
Against Claude Haiku 4.5: GPT-6 Astra answered 1.2× later (2.6 s to 2.2 s), cost 5.6× more, and passed 5 of 5 to its 5.
Against GPT-6.1 Sol: GPT-6 Astra answered 1.5× later (2.6 s to 1.8 s), cost 9.9× more, and passed 5 of 5 to its 5.
Against GPT-6 Sol: GPT-6 Astra answered 1.2× later (2.6 s to 2.2 s), cost 4.7× more, and passed 5 of 5 to its 5.
Against GPT-6 Luna: GPT-6 Astra answered 1.2× later (2.6 s to 2.2 s), cost 86.8× more, and passed 5 of 5 to its 4. Only GPT-6 Astra passed Pens at 3 for $4.
Against Gemini 4 Argon: GPT-6 Astra answered 1.7× sooner (2.6 s to 4.4 s), cost 5.5× more, and passed 5 of 5 to its 5.
Against Gemini 3.1 Pro (preview): GPT-6 Astra answered 3.2× sooner (2.6 s to 8.3 s), cost 1.2× less, and passed 5 of 5 to its 5.
Against Gemini 3.8 Flash: GPT-6 Astra answered 1.8× sooner (2.6 s to 4.7 s), cost 6.0× more, and passed 5 of 5 to its 5.
Against DeepSeek V4.1 Flash: GPT-6 Astra answered 2.8× later (2.6 s to 0.9 s), cost 34.2× more, and passed 5 of 5 to its 5.
Against DeepSeek V4 Pro: GPT-6 Astra answered within 10% of its time (2.6 s to 2.7 s), cost 9.7× more, and passed 5 of 5 to its 5.
Against Grok 4.7: GPT-6 Astra answered 4.5× sooner (2.6 s to 11.7 s), cost 1.2× more, and passed 5 of 5 to its 5.
Against Kimi K3: GPT-6 Astra answered 2.1× sooner (2.6 s to 5.4 s), cost 2.3× more, and passed 5 of 5 to its 5.
Against GLM 5.3: GPT-6 Astra answered 2.8× later (2.6 s to 0.9 s), cost 24.2× more, and passed 5 of 5 to its 5.
Against GLM 5.3 Flash: GPT-6 Astra answered within 10% of its time (2.6 s to 2.7 s), cost 56.9× more, and passed 5 of 5 to its 5.
Against Mistral Large 4: GPT-6 Astra answered 2.0× sooner (2.6 s to 5.1 s), cost 4.7× more, and passed 5 of 5 to its 5.

How these runs were done

Final answer. Automatic. The reply's last “Final answer:” line must hold the right value.

How the runs were done, and every math prompt.

GPT-6 Astra, and math, elsewhere

Questions

Is GPT-6 Astra good for math?

In our test runs it passed 5 of 5 math prompts, tied 1st of the 20 models in llmwise. Every reply is on this page, so you can judge them yourself.

How many of my messages does a math reply from GPT-6 Astra use?

1 message each on Pro, where it has 31 a month on Pro. The price of a message is fixed and shown before you send it, however long the reply.

How were these runs done?

The same way for every model: each prompt sent through llmwise's own pipeline, each reply checked the same way. The methods page has every prompt and how each is scored.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.