◂ BACK TO LEADERBOARD
▚ MODEL AUDIT // GGUF

// Muse Glimmer 28B

FLAWS — Minor Logic Flaws 88/100 tests: PASS
⚠ SPEED CAVEAT: Slow deep-thinker: ~17-19 t/s, 5-9 min/prompt, ~5-9k tokens incl. reasoning. Fast for quality, not throughput.
tok/sec17.17
tokens7308
TTFT3.84s

▮ PILLAR BREAKDOWN

Query Safety
18/20
Pooling
19/20
Transactions
17/20
Pagination
19/20
Test Integrity
15/20

✓ WENT RIGHT

  • Pooling (19/20)
  • Pagination (19/20)
  • Query Safety (18/20)
  • Transactions (17/20)

✗ WENT WRONG

  • No pillar fell below 14 — solid across the board.

▮ CRITICAL BUGS

▮ RECOMMENDED USE

Clean parameterized queries throughout (zero interpolation, with a real injection-attempt test storing DROP TABLE as a literal label), semaphore-pooled concurrency with 100-call leak proof, proven pagination (true COUNT, last-page remainder, out-of-range=empty), and snapshot-based rollback verified on partial failure. Edges out Gemma-26B's 86 on the same prompt.

▮ REFACTORED PATCH

# 1. Add forced-timeout test (fills the test_integrity gap)
async def test_acquire_timeout():
    pool = MockPool(max_size=2, acquire_timeout=0.1)
    svc = UserService(pool)
    hold = [await pool.acquire() for _ in range(2)]  # exhaust
    try:
        await asyncio.wait_for(svc.get_users(1, 10), timeout=1.0)
        assert False, 'should have timed out'
    except (TimeoutError, asyncio.TimeoutError): pass
    finally:
        for c in hold: await c.release()
    assert pool.checked_out == 0

# 2. Strengthen rollback — track applied writes and undo only those
applied = []
try:
    for uid, label in pairs:
        await conn.execute('UPDATE users SET label = $1 WHERE id = $2', [label, uid])
        applied.append(uid)
    return len(pairs)
except Exception:
    for uid in reversed(applied): MockDB.users[uid]['label'] = snapshot[uid]
    return 0