> MTP GGUF version here, MLX version here, and MLX MTP version here. > ⚠️ **WARNING - Read before use**: Abliterated models like CyberTiel are able to say and do things other models refuse, including potentially harmful behaviours. By using CyberTiel, you agree to take full personal responsibility and liability for your use of it, its behaviour and generated content, and to show caution: it is entirely up to you as the user to ensure your use of CyberTiel is legitimate, legal and harmless, and that the model is safely sandboxed and monitored when running. Much like a knife, abliterated models like CyberTiel can be classified and used as either a tool or a weapon, depending on the context and use case. We carry forward huihui's original usage warnings. *CyberTiel* outcodes every other 35B-A3B at Q4 quantization (and spring-of-2026 frontier models), while engaging with offensive security work without hesitation or refusal. As a sweet spot between speed and ability, CyberTiel delivers agentic coding solves about 3-4x faster than 3.8-27B dense. This is the first time the frontier coder in this size/speed class is an uncensored model. If you need a safer censored alternative, go for *TielCoder*.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗> GGUF without MTP here, MLX version here, and MLX MTP version here. > ⚠️ **WARNING - Read before use**: Abliterated models like CyberTiel are able to say and do things other models refuse, including potentially harmful behaviours. By using CyberTiel, you agree to take full personal responsibility and liability for your use of it, its behaviour and generated content, and to show caution: it is entirely up to you as the user to ensure your use of CyberTiel is legitimate, legal and harmless, and that the model is safely sandboxed and monitored when running. Much like a knife, abliterated models like CyberTiel can be classified and used as either a tool or a weapon, depending on the context and use case. We carry forward huihui's original usage warnings. *CyberTiel* outcodes every other 35B-A3B at Q4 quantization (and spring-of-2026 frontier models), while engaging with offensive security work without hesitation or refusal. As a sweet spot between speed and ability, CyberTiel delivers agentic coding solves about 3-4x faster than 3.8-27B dense. This is the first time the frontier coder in this size/speed class is an uncensored model. If you need a safer censored alternative, go for *TielCoder*.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗With our **Sharp chat template, MTP, and vision baked in**, the model answers lean and stays on-task out of the box. No template wrangling: download, point llama.cpp at it, go. If you want it to think deeper, set the effort level through `chat_template_kwargs`: Levels: `low`, `medium`, `xhigh` — `high` is accepted but is an alias for `xhigh`, not a step below it. Omit it for Dirk's lean default (medium). Turn thinking off entirely with `"enable_thinking": false`. - **Base:** `Qwen/Qwen3.8-27B`, a dense 27B **vision-language** model (vision preserved). - **Quant:** two quantizers, each where it is strongest. From 3 bpw up, Unsloth's **Dynamic 3.0** **UD** quants — their current generation, not an older ladder. Below that, **GSQ-RCO** quants from IST-DASLab, which hold up substantially better at 2–3 bpw (see the note under the file table). Every tier **keeps the model's MTP (`nextn`) head**: runtimes with multi-token-prediction speculative decoding can use it for faster generation. - **Template:** the Sharp chat template (Qwen 3.8-aware) — froggeric's fixed Qwen template plus a terseness system prompt, and turning off the xhigh thinking default. Every tier supports the terseness opt-out (`chat_template_kwargs: {"terse": false}`). The `GSQ-RCO-` tiers carry `v22.4.1`, which also stands down when the runtime injects its own tool protocol (an LM Studio fix); the `UD-` tiers carry `v22.4.0` and are otherwise identical — they pick up `v22.4.1` on the next pass. It is byte-swapped into the GGUF metadata; the weights and the MTP tensors are untouched.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗*Nail* dominates Qwen3.6-27b and its ThinkingCap fine-tune on time-to-answer while matching their accuracy on reasoning and agentic software engineering ability, and wins on multi-turn conversation quality *even when the 27b models use a higher quantization*. > **This is Unsloth's UD-Q4_K_XL quant of Qwen3.6-35B-A3B with an improved chat template also containing a force-appended system prompt that sharpens behaviour.** llama.cpp applies it automatically, nothing needed from you. > >If you have a Mac, use the MLX build instead. This exists for everyone who doesn't. Best conversation score, fastest to a working fix, fastest to a correct answer — against dense models that take 2-5x times as long, at accuracy parity with all of them. Four benchmarks, n=3 seeds, one Mac Studio M2 Ultra. Nail beats Opus4.8 (medium) 3-0 on first-attempt solves in a real SWE Live repo, no fluffing about. We used Pi for local models, Claude Code for Opus.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗*Nail* dominates Qwen3.6-27b and its ThinkingCap fine-tune on time-to-answer while matching their accuracy on reasoning and agentic software engineering ability, and wins on multi-turn conversation quality *even when the 27b models use a higher quantization*. > **These are Unsloth's MTP GGUF quants of Qwen3.6-35B-A3B with our improved chat template and force-appended system prompt baked in** — the identical recipe to the standard Nail GGUF, only this build keeps the **multi-token-prediction (MTP / `nextn`) tensors** that the standard quants drop. Runtimes that support MTP speculative decoding can use them for faster generation. llama.cpp applies the template automatically, nothing needed from you. > > **The benchmarks below were run on the non-MTP models.** MTP adds a draft head for speculative decoding — accuracy is unchanged either way. Unsloth reports ~1.5–2× faster inference in vLLM/SGLang. In llama.cpp it needs a recent source build (b10362+) run with `--spec-type draft-mtp`; we measured **~1.35× on code generation** tuned (`--spec-draft-n-max 5 --spec-draft-p-min 0.6`), ~1.25× at defaults (M2 Ultra, Q4). Current llama.cpp *releases* don't ship MTP yet, so there it's parity with the standard GGUF. > >If you have a Mac, you can use the MLX build. Best conversation score, fastest to a working fix, fastest to a correct answer — against dense models that take 2-5x times as long, at accuracy parity with all of them. Four benchmarks, n=3 seeds, one Mac Studio M2 Ultra. Nail beats Opus4.8 (medium) 3-0 on first-attempt solves in a real SWE Live repo, no fluffing about. We used Pi for local models, Claude Code for Opus.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Dynamic **imatrix** quants of openbmb/MiniCPM5-2B, a 2.5B dense 131k-context model, carrying our cyber-and-coding-weighted importance matrix, an unsloth-dynamic-style per-tensor quant policy, and the **Sharp-MiniCPM** chat template with fixes and an improved system prompt. We recommend the Q6_K_XL for anyone who can run it. Practically lossless, at under half the size of the F16 model, and scoring higher than stock Q8_0 on SWE-bench-Live. **99% KL** is the 99th percentile of KL divergence against the BF16 weights on held-out code — the rare token where a quant actually diverges. **top-1** is how often the quant picks the same next token as BF16. The last three columns are the largest `q8_0` KV context that fits beside the weights on a 4 / 6 / 8 GB card.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Dynamic **imatrix** GGUF quants of XHToken/Spark-X2.5-4B, a 4B dense long-context model, carrying our cyber-and-coding-weighted importance matrix, a heuristic per-tensor bit allocation, and the **Sharp-Spark** chat template. This is a **small, fast, long-context coder that fits 6 GB-VRAM GPUs** and still holds usable speed out to six-figure context. If you have more regular RAM than 16GB, try a (Cyber)TielCoder with partial GPU offloading instead. Its MoE architecture makes it fast even when it's split onto system RAM, and its about twice as capable.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗*Tiel* is the fast coder of the arsenal. At 4-bit quantization and 22 GB it fixes real codebase issues at the rate (and speed, with the right GPU) of Opus 4.6 medium, while holding the best multi-turn conversation of any local model we have measured. It is also cheerfully bad at trivia. > **This is Ornith-1.5-35B-A3B re-quantized > dynamically with our own imatrix and carrying the Sharp chat template** > inside the GGUF. Find TielCoder MTP GGUFs here. > **The uncensored and improved CyberTiel is out now!** Get it here. **Where it stands.** On 25 SWE-bench-Live problems Tiel fixes 12 — the same as Opus 4.6 (medium), four more than Ornith-1.5 itself, three more than Nail, and four more than Sonnet 5 (medium). Among models of its own class it is first; the ones ahead are dense 27Bs and Opus 5. Its time per attempt is also steadier than Nail's: an 8.6 minute median against 7.2, but a 12.3 minute mean against 15.7, because it lacks Nail's tail of expensive attempts.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗*Tiel* is the fast coder of the arsenal. At 4-bit quantization and 23 GB it fixes real codebase issues at the rate (and speed, with the right GPU) of Opus 4.6 medium, while holding the best multi-turn conversation of any local model we have measured. It is also cheerfully bad at trivia. > **This is Ornith-1.5-35B-A3B re-quantized > dynamically with our own imatrix and carrying the Sharp chat template** > inside the GGUF. Find the non-MTP GGUFs here, and the non-MTP MLX > version here. > **The uncensored and improved CyberTiel is out now!** Get it here. **Where it stands.** On 25 SWE-bench-Live problems Tiel fixes 12 — the same as Opus 4.6 (medium), four more than Ornith-1.5 itself, three more than Nail, and four more than Sonnet 5 (medium). Among models of its own class it is first; the ones ahead are dense 27Bs and Opus 5. Its time per attempt is also steadier than Nail's: an 8.6 minute median against 7.2, but a 12.3 minute mean against 15.7, because it lacks Nail's tail of expensive attempts.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗An imatrix-calibrated, **Unsloth-Dynamic-style** GGUF ladder for **Ornith-1.5-35B-A3B** — the model's own weights and chat template, re-quantized with a per-tensor precision policy and a fresh importance matrix. **The MTP (`nextn`) block is removed from every tier** — Ornith-1.5 ships it untrained, so it could only add size. Why. > ### Not affiliated with Unsloth > This is an independent, best-effort community reproduction of Unsloth's > Dynamic GGUF method — **not** an > official Unsloth release, and not endorsed by or connected to Unsloth in any way. Unsloth had not > published their own Ornith-1.5 quants at the time of writing and a lot of people wanted them, so > we published these while we wait. If and when Unsloth ship an official build, use theirs. "Unsloth" > is used here only to name the *method* these quants reproduce. 1. **Unsloth-Dynamic per-tensor policy.** Instead of one bit-width for the whole model, each tensor gets a type chosen by sensitivity — more bits where they matter (embeddings, attention, early/late layers), fewer where they don't. We transfer the published per-tensor recipe from Unsloth's `Qwen3.6-35B-A3B-GGUF` (same `qwen35moe` architecture) tier-for-tier. 2. **A fresh importance matrix.** Computed on Ornith's own `BF16` over an interleaved multilingual + code calibration corpus, so the quantizer knows which weights to protect for *this* model rather than inheriting someone else's statistics.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗