Tiel-Coder 35B-A3B has 34.661 billion total parameters. The selected UD-Q4_K_XL GGUF file is 22.36 GB.
Data snapshot: . 3 published tests in this catalog.
Model file and details
Provider
peculiar-ragdoll
Total parameters
34,660,610,688
Exact file size
22,360,478,080 bytes
Quantization / precision
UD-Q4_K_XL
Weight repository
peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF
Original model repository
peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF
Weight revision
a1ae34c609d9ef03827f373a215795f22545334c
Model revision
a1ae34c609d9ef03827f373a215795f22545334c
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-08-19
Date evidence
The MIT-licensed GGUF repository was created on 19 August 2026; the pinned revision embeds the Sharp chat template and a code-weighted importance-matrix Dynamic quantization of Ornith-1.5-35B-A3B.
These scores were compiled from the linked sources. These published scores were not independently tested here. Different prompts, precision, thinking budgets and tools can change results.
MMLU-Pro
73.7 / 100
Knowledge and reasoning across many subjects. Published accuracy; different evaluation settings can change the result.
Benchmark efficiency: 3.296 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher MMLU-Pro evaluation of the exact benchmarked 4-bit build across three seeds. The card compares the Sharp-template Tiel build with a same-quant Ornith control.
Publisher-run result, not independently reproduced. The card does not state the full prompt format, evaluator version, aggregation method or hardware; small differences should be treated as noise. This score documents a known trade-off rather than a coding-quality claim.
Publisher run on 25 selected SWE-bench-Live issues, one attempt per issue. A small coding-agent sample; issue selection, scaffold, runtime and hardware materially affect results.
Benchmark efficiency: 2.147 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher evaluation of the exact 22.36 GB UD-Q4_K_XL tier on 25 selected SWE-bench-Live issues, one run per issue. The card reports 12 solved issues, an 8.6-minute median and a 12.3-minute mean per attempt.
Publisher-run 25-issue subset, not the SWE-bench Verified benchmark and not an independent evaluation. The card does not fully specify issue selection, agent scaffold, runtime build, hardware or every generation setting, so the percentage is isolated in a limited benchmark and must not be compared with SWE-bench Verified scores.
Publisher multi-turn score across 114 conversations. Answer quality is weighted four to one against clarifying questions; not a general benchmark or independent evaluation.
Benchmark efficiency: 3.005 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher Claw-Eval multi-turn evaluation covering 114 scored conversations for Tiel-Coder. The score weights answer quality four to one relative to clarifying questions.
Publisher-run specialist result for the benchmarked 4-bit build. The card does not provide a complete reproducible harness, task inventory, runtime or hardware specification, and no independent rerun was found.
Context and version evidence: The repository and file name identify the selected model. Revision hashes are shown above when recorded. Tested context length, full hardware configuration and run date are not consistently recorded in this snapshot. Only settings explicitly stated in each benchmark record are known.
The 22.36 GB figure covers model weights in this file. It excludes the context cache, runtime, operating system and any separate multimodal files. It is not a measured RAM requirement or a token-generation speed.
These complete configurations pass the initial weight-size check for the selected 22.36 GB file. Allow extra memory for context and software. Check the model’s runtime notes and your inference software before buying.
Listing evidence: 2026-09-25. Amazon.com base price visible in an owner-supplied screenshot reviewed 25 Sep 2026; capture time not supplied. Dated reference, not a live quote. Excludes card signup offers, coupons, tax and shipping.
Listing evidence: 2026-10-02. Apple US launch price for the base M5 Max Mac Studio, published 22 September 2026: USD 2,499. Dated manufacturer reference, not an Amazon offer; tax and shipping may differ.
Listing evidence: 2026-09-24. GMKtec, EVO-X2 / 64GB / 1TB, recorded 22 Sep 2026; exact ASIN capacity identified 24 Sep. This is a dated reference, not a current Amazon price; tax, shipping and bundle contents may differ.
Memory matches are estimates, not tested compatibility or speed. As an Amazon Associate I earn from qualifying purchases.
Confirm model architecture and quantization support in your runner. The selected weights use 22.36 GB; this excludes context cache, runtime buffers and other programs.
These are the first listed tiers that pass our reserve policy, not measured minimum requirements. System RAM and VRAM remain separate. Use the memory tool to change the reserve for your context and runner.