Skip to content
CrucibleMark

Occamy 1.0 35B-A3B (Accio-Lab)

Occamy 1.0 by Accio-Lab is an agentic derivative of the Qwen3.6-35B-A3B checkpoint, focused on long-horizon co-work sessions with tools, structured APIs, and persistent state tracking. The NVFP4 quantization is selective: only the routed experts are quantized, while attention, router, embeddings, and output head remain in BF16. The 35-billion-parameter MoE activates only 3 billion parameters per token and supports 262,000 tokens of context. Apache 2.0 license and documented provenance with recipe, data, and validation artifacts.

Accio-Lab Version 1.0 Commercial use permitted MoE 35 B (3 B active) 262 K Context 12/2025 locally tested

  • Open Weights
  • Workstation
  • vLLM
  • Text
  • Vision
  • Interactive

Sovereign Risk: MEDIUM Occamy 1.0 NVFP4 is a community derivative of Qwen/Qwen3.6-35B-A3B, with a published provenance trail including the quantization recipe, data-provenance file, and validation artifacts. The upstream base is Apache 2.0, the checkpoint runs locally, and the NVFP4 export is limited to routed experts, but Accio-Lab’s organizational jurisdiction is not publicly documented, so the provenance risk remains medium.

Tool-use profile: 6 assets in detail

Comparison of asset performance (P1/P2/Combined) against the fleet average

Asset performance (radar)

Score Breakdown vs. Fleet Average

Tool-use details

Asset performance, reliability, and runtime profile

Name P1 P2 Combined
EU License Research 95 60 74
HTTP Fetch & Extract 80 80 80
Tool Failure Handling (404) 80 80 80
Web Search & Tool Selection 35 20 28.3
URL Construction & Fetch 80 80 80
Multilingual Search & Synthesis 100 40 70

CrucibleMark evaluates tool use across 6 independent tests. Click a test name for details.

Reliability

  • Tool Call Valid: No
  • Retry: Not required
  • Hallucination: Not detected

Reliability measures how consistently a model actually executes tool calls: Tool Call Valid schema and format accepted, Retry Required successful only after retry, Hallucination Flag fabricated tools or parameters detected. All three green means production-ready.

Runtime profile

Call 1
3.04
First Request
MCP
1.2
Protocol Latency
Synthesis
11.85
Response Generation
Total
96.52
Sum of All Phases
Token
10126
Input + Output
Cost
$0
Cost per Run

The runtime profile shows the latency and cost metrics for the model run: Call 1 First Request, MCP Protocol Latency, Synthesis Response Generation, Total sum of all phases, supplemented by Token input and output and Cost cost per run.

Tool-use review

Created

Deployment Verdict

Conditional deploy, because tool execution is mostly sound, but tool-call validity is inconsistent and overall synthesis is not stable enough for autonomous MCP pipelines.

Tool Execution Profile

Occamy 1.0 35B-A3B shows usable operational tool usage, but no reliable tool intelligence. The strong signal comes from clearly guided retrieval paths: in the HTTP Fetch & Extract test it extracts structured facts cleanly, and in the URL Construction test — which checks whether a target URL can be derived from prior knowledge — it works with solid precision. Multilingual research also starts out operationally strong.

The weak point is selecting the right tool without a hint. In the Web Search & Tool Selection test, which checks whether web_search is needed before fetch, it falls off noticeably. This argues against flexible tool planning and suggests a pattern instead: when a URL or direct retrieval path appears plausible, the model follows that path rather than correctly classifying the information need first. For MCP orchestration this means: usable in tightly framed flows, risky in open research paths with multiple possible tools. The fact that tool calls were not consistently valid sharpens exactly this point.

Synthesis Fidelity

How well does it consolidate tool results? Only to a limited degree. The P2 performance shows a consistent pattern: solid on direct extraction tasks, but weak as soon as multiple sources, language switches, or implicit inferences need to be combined. Particularly striking is the gap between perfect operational execution on Multilingual Search & Synthesis and noticeably weaker consolidation in German. The model finds material, but does not compress it precisely enough for reliable decision outputs.

Does it stay within tool results or fall back on training data? Mostly yes, and that is the most important positive finding. In the Honeypot EU License Research test — which checks whether current license restrictions are actually retrieved from web sources — Occamy does not hallucinate. The trust foundation is therefore present: it does not fabricate compliance facts, even when the summary is only moderately strong.

Error Resilience

The model responds in a production-ready manner to tool failures. In the 404 test, which measures transparent error handling rather than fabricated page content, it communicates the failure cleanly and does not hallucinate substitute content. This is acceptable for production pipelines. A system can handle clear errors; it cannot handle fabricated results.

Sovereignty Profile

Locally deployable and, despite a local_sovereign setup, narrowly fleet-competitive. At 68.71 Combined, it sits 0.68 points above the fleet average of 68.03.

Conclusion & Recommendation

Suitable for local, sovereign pipelines with pre-structured tool paths, explicit fetch steps, and human review at the output level. Not suitable as an autonomous research orchestrator that must decide on its own whether to search, fetch, or replan. If you hard-code tool selection and call schema, Occamy can be a useful local worker. If the model is expected to navigate the infrastructure independently, the risk remains too high.

This analysis was generated automatically based on benchmark data. Model used: GPT-5.4 by OpenAI. The raw data and full methodology are documented in the GitHub project.