◂ BACK TO LEADERBOARD
▚ MODEL AUDIT // n/a (cloud)

// DeepSeek V4 Flash (CLOUD baseline) CLOUD BASELINE

PROD — Production Ready 91/100 tests: PASS
⚠ SPEED CAVEAT: Cloud model (run via opencode, not LM Studio) — tok_sec/tokens/TTFT are N/A (not measured for cloud). Included as a quality baseline against the local models. NOTE: it took 3 attempts to produce any output and ~12 minutes of thinking before succeeding — so it is a QUALITY benchmark, not a speed/usability one.
runtimeCLOUD — not measured

▮ PILLAR BREAKDOWN

Complexity (O(1))
18/20
Concurrency / Races
18/20
Tx Isolation
19/20
Memory & Edges
18/20
Test Integrity
18/20

✓ WENT RIGHT

  • Tx Isolation (19/20)
  • Complexity (O(1)) (18/20)
  • Concurrency / Races (18/20)
  • Memory & Edges (18/20)
  • Test Integrity (18/20)

✗ WENT WRONG

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

▮ CRITICAL BUGS

▮ RECOMMENDED USE

Reference-quality baseline (91/100) — the bar the local models are measured against. Only submission with __slots__ + time.monotonic() + delta-based transactional frequency accounting. Sets the ceiling for correctness, though its unreliability (3 attempts, 12-min think time) makes it a poor *local* daily-driver.

▮ REFACTORED PATCH

# These are minor refinements on an already production-ready file.

# FIX 1: eliminate the recovery min() scans by keeping min_freq
# strictly in sync on every insert/bump/remove (it already does on
# the hot path), so the _evict_one recovery branch is unreachable and
# can assert instead of scanning:
assert self._min_freq in self._freq_to_list or not self._freq_to_list

# FIX 2: extend __slots__ to Transaction and LFUCache.
class LFUCache(Generic[KT, VT]):
    __slots__ = ('_capacity','_key_to_node','_freq_to_list','_min_freq',
                 '_lock','_ttl_index','_evictor_task','_closed')

# FIX 3 (optional): on commit, if a key's global node changed since the
# tx snapshot, raise LFUCacheError('lost update') instead of overwriting.