Laguna-S-2.1 Local Test: Fastest 100B+ Model but Fails Grounding Under Pressure
klinec · reddit · 2026-07-22
A developer put the newly released Laguna-S-2.1 through a deep private agentic eval on a single RTX Pro 6000 (96GB), comparing it against Qwen3.5-122B and others.
Strengths:
- Top-tier tool calling & speed: Best tool-arg selection (0.89 pass rate) and handles tool chains up to 6 levels deep (Qwen manages 4).
- Fastest 100B+: Hits 109 tok/s at 256k ctx, the fastest 100B+ model tested on this card.
- Rock-solid formatting: Zero JSON/streaming errors; recovers from validation errors on the first retry.
Critical Flaws:
- Hallucinations under pressure: Disqualifying grounding behavior with 3 hand-confirmed hard fabrications (invented P&L and named nonexistent horses), whereas Qwen had zero inventions across 240 runs.
- Lacks breadth: Sports knowledge and odds math score significantly lower than Qwen (0.80 vs 1.00).
Deployment Gotchas:
- Requires maxtokens 8k+; otherwise, it burns the entire budget thinking and returns empty.
- dflash speculative decoding boosts single-stream speed massively (up to 271 tok/s) but causes a net loss under concurrent streams and lacks bitwise stability at temp 0.
Verdict: An excellent agentic coding specialist with superb mechanics, but its tendency to confidently invent facts when data runs out makes it unsuitable for fully autonomous general use.
More from coding & agent
- Dev builds interactive 3D product experience with GPT-6 Astra + Hyper3D Rodin — nikola_mr64990 · 2026-09-11
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- First-ever Three.js Conference lands in Paris, with a panel on AI-shortened design workflows — OdinLovis · 2026-09-11
- Data engineering, not agent frameworks, is the real bottleneck for enterprise AI agents — dhruv2038 · 2026-09-11