SQL · 6 checks · max 500 tokens · benched 2026-09-16

Muse Spark 1.3 vs Muse Glimmer 30B vs Llama 4 Maverick vs Llama 4 Scout on top five customers by 2025 spend

Meta models side by side on "Top five customers by 2025 spend": Muse Glimmer 30B scores 10/10; Llama 4 Scout is the cheapest answer scoring 8+ at $0.03 per 1,000 runs. Outputs, checks, judge reasons, latency and cost.

The prompt every model received

System

You are a database engineer. Reply with a single SQL statement and nothing else.

User

Tables: customers(id, name) and orders(id, customer_id, amount, ordered_at). Write one standard SQL query that returns each customer's name and their total order amount for calendar year 2025, highest total first, limited to the top 5. Use a JOIN, GROUP BY and ORDER BY.

Rubric for the judge: Correct join and aggregate, filters ordered_at to 2025, sorts descending, limits to 5, portable SQL.

Side by side

Every cell is one OpenRouter call at temperature 0 with the prompt's token cap and reasoning effort "low" where the model supports it. Cost is usage × the catalogue rate in models.json. Quality is one judge call to anthropic/claude-haiku-4.5 against the prompt's rubric, cached per prompt version.

Muse Spark 1.3

meta/muse-spark-1.3

unjudged

failed: OpenRouter HTTP 403: This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.

Latency: 0ms
Cost: $0
Per 1,000: $0

0 in · 0 out · 0 words · checks 0/6

regex
regex
regex
regex
regex
contains

(empty output)

Judge: —

Muse Glimmer 30B

meta/muse-glimmer-30b

10/10

top judge score

Latency: 109ms
Cost: $0.00034
Per 1,000: $0.34

105 in · 282 out (195 reasoning) · 37 words · checks 6/6

regex
regex
regex
regex
regex
contains

```sql
SELECT c.name, SUM(o.amount) AS total_order_amount
FROM customers c
JOIN orders o ON c.id = o.customer_id
WHERE o.ordered_at >= DATE '2025-01-01' AND o.ordered_at < DATE '2026-01-01'
GROUP BY c.id, c.name
ORDER BY total_order_amount DESC
LIMIT 5;
```

Judge: Query correctly joins customers and orders, filters 2025 using standard DATE literals, groups by customer with proper aggregation, orders descending by total, limits to 5, and uses portable SQL syntax.

Llama 4 Maverick

meta-llama/llama-4-maverick

9/10

Latency: 219ms
Cost: $6.2e-5
Per 1,000: $0.06

93 in · 63 out · 32 words · checks 6/6

regex
regex
regex
regex
regex
contains

```sql
SELECT c.name, SUM(o.amount) AS total_amount 
FROM customers c 
JOIN orders o ON c.id = o.customer_id 
WHERE EXTRACT(YEAR FROM o.ordered_at) = 2025 
GROUP BY c.name 
ORDER BY total_amount DESC 
LIMIT 5;
```

Judge: Query correctly uses JOIN, GROUP BY, ORDER BY, and LIMIT with proper 2025 filtering; minor deduction because GROUP BY should include c.id to be fully standard-compliant (though c.name alone works in most systems).

Llama 4 Scout

meta-llama/llama-4-scout

9/10

cheapest scoring 8+

Latency: 689ms
Cost: $2.8e-5
Per 1,000: $0.03

93 in · 63 out · 32 words · checks 6/6

regex
regex
regex
regex
regex
contains

```sql
SELECT c.name, SUM(o.amount) AS total_amount
FROM customers c
JOIN orders o ON c.id = o.customer_id
WHERE EXTRACT(YEAR FROM o.ordered_at) = 2025
GROUP BY c.name
ORDER BY total_amount DESC
LIMIT 5;
```

Frequently asked

What does this prompt test?

SQL: Correct join and aggregate, filters ordered_at to 2025, sorts descending, limits to 5, portable SQL. The deterministic checks are regex, regex, regex, regex, regex, contains.

Which model should I pick for this task?

If the judge's bar of 8/10 is good enough for you, Llama 4 Scout at $0.03 per 1,000 runs. If you need the top score, Muse Glimmer 30B at $0.34 per 1,000 runs.

Same prompt, other families

Run it yourself

Re-run this prompt on Meta with your own OpenRouter key, or tweak the wording and see what changes.