HERETIC 1.4.0 MTP APEX MIT Ornith-1.5-35B-A3B-Heretic-MTP-APEX English | 📖 中文文档 Heretic Trial 62 BF16 merge · replaced MTP head · APEX GGUFs + mmproj 🐦 About this release Ornith-1.5-35B-A3B is DeepReinforce's Qwen3.5 MoE agentic coding model. This experimental derivative merges the selected Heretic 1.4.0 Trial 62 LoRA into BF16 weights. The shipped native MTP head is untrained (initializer-like statistics). It is replaced here with the 19-tensor fused BF16 head from shisa-ai/Ornith-1.5-35B-A3B-MTP-ONLY (Qwen3.6 graft + 12K KL distillation). Refusal-modified models can behave differently from the original; evaluate before deployment. License: MIT. 🧠 Model Details Architecture Qwen3.5 MoE, multimodal Parameters 35B total, 3B active per token · 256 experts, 8 active Layers 40 transformer layers + 1 fused MTP layer (19 tensors, Q8_0 in APEX) Context 262,144 tokens License MIT
Explore AI models 41
Ornith-1.5-35B-A3B-MTP-APEX English | 📖 中文文档 Self-improving agentic coding model · APEX quantized GGUFs + BF16 + mmproj Ornith-1.5-35B-A3B is a self-improving agentic coding model from the Ornith Team . It extends Ornith-1.0 by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts — continuously generating new training tasks, discovering effective strategies, and improving the policy through reinforcement learning. Activating only ~3B parameters per token, it significantly outperforms its similar-sized peer Qwen3.6-35B across all coding and agentic benchmarks: Terminal-Bench 2.1 67.8 , SWE-bench Verified 79 , SWE-bench Pro 59.6 , SWE-bench Multilingual 71.4 , NL2Repo 46.2 , MCP-Atlas 70.2 , ClawEval 72.5 . This GGUF package includes the mmproj-BF16.gguf vision projector for multimodal (image + text) capabilities with llama.cpp. Unlike Ornith-1.0, the MTP layer is native to the model — no external grafting required. License: MIT. Architecture Qwen3.5 MoE (Mixture of Experts) Parameters 35B total, 3B active per token Experts 256 routed experts, 8 active per token Layers 40 transformer layers + 1 MTP layer Context 262,144 tokens MTP 1 native MTP layer (785 tensors) License MIT
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
Qwen-AgentWorld-35B-A3B-MTP-Uncensored-APEX English | Chinese README AgentWorld world model · Abliterix uncensored · native MTP injected · APEX-I GGUFs Qwen-AgentWorld is designed as an agentic environment simulation / language world model. It is intended to simulate agent-environment interactions and predict environment states from actions and history. It is not recommended as a general daily assistant, coding model, or autonomous Agent backbone unless you specifically need this world-model behavior. This package contains GGUF releases of Qwen-AgentWorld-35B-A3B , a Qwen3.5 MoE world model for agentic environment simulation. The model was modified with an Abliterix uncensored LoRA, stream-merged into safetensors, injected with one native MTP layer from a compatible Qwen3.5-35B-A3B MTP shard, converted to BF16 GGUF, and quantized into APEX-I tiers using a model-specific imatrix. The release keeps the BF16 GGUF, imatrix, and mmproj-F16.gguf for reproducibility and multimodal llama.cpp use.
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
Optimized quantization strategy with improved imatrix calibration. File sizes reduced while maintaining quality: I-Compact: 16.61 GB → 15.85 GB (-0.76 GB) I-Balanced: 24.24 GB → 24.18 GB (-0.06 GB) I-Quality: 21.90 GB → 21.90 GB (unchanged) These GGUF files are quantized using APEX , a novel MoE-aware mixed-precision quantization technique that outperforms standard quantization methods while being significantly smaller. APEX beats Q8_0 perplexity at half the size — and even beats F16. APEX classifies every tensor by its role — routed expert, shared expert, or attention — and applies a layer-wise precision gradient, giving the most sensitive edge layers higher precision and compressing the redundant middle layers more aggressively.
FIT-GGUF v0.1 ZERO-BYTE VERIFIED 14 TIERS · 7–13.5 GiB MTP HEAD REMOVED MEASURED KL + SAME-TOP APACHE-2.0 Qwen3.8-27B-Uncensored · FIT-GGUF Fourteen continuous-size GGUF quantizations — every file exactly its predicted size, every tier measured. 🧭 About FIT-GGUF — the tool behind these files Every file in this repository was planned, executed and verified by FIT-GGUF , an open-source, deterministic tensor-level planning layer on top of llama.cpp quantization presets. Standard GGUF quantization asks you to pick one of a handful of presets; FIT-GGUF starts from the largest supported preset below your requested byte budget, then spends the remaining bytes on deterministic tensor-level precision upgrades. Traditional GGUF gives you presets. FIT gives you a size slider. Deterministic size prediction & recipe execution ✅ Validated Universally optimal tensor allocation ⚠️ Not established — FIT claims precise size control, not a universal quality optimum The method, the full preregistered research record and the fit CLI are open source: github.com/Scorp1o117/FIT-GGUF ⚠️ Safety notice / 安全提示 The source model is an abliterated, refusal-removed model with no meaningful built-in guardrails, and may comply with harmful, illegal or unsafe requests. Use it only where you can provide appropriate moderation, access control and legal review. Do not deploy it to end users without your own safety layer. 源模型经过拒答方向移除,不具备可靠的内置安全护栏;请仅在合法、受控、具备审核与滥用防护的环境中使用,使用者自行承担部署责任。
GSQ-RCO upstream · layout preserved 144 tensors transplanted IQ3_S / IQ3_XXS / IQ2_XS / Q2_0 per-tensor blake2b verified 262K context APACHE-2.0 Qwen3.8-Flash-Next · GSQ-RCO-abliterated The refusal direction removed from an already-quantized GSQ-RCO model — 144 tensors across all 48 layers swapped for ready-made abliterated weights, with every value and scale GSQ learned left untouched and the upstream per-tensor type assignment restored exactly . The IQ3_S build is verified tensor by tensor: everything that changed was supposed to, nothing else moved. 38.0 GB · Q2_0 every tier costs only 0.25 – 0.44 GB over upstream 55.1 GB · IQ3_S 🧭 What this repository is An abliterated build of the official ISTA-DASLab GSQ-RCO quantizations . Not a single one of GSQ's learned quantized tensors was recomputed — 144 "write-to-residual-stream" projection tensors across all 48 layers were replaced with the corresponding tensors from a ready-made abliterated release. Base ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF — GSQ learned quantization + RCO budget-constrained type assignment Ablation source orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF — the 144 target tensors are taken straight from it Method Byte-level GGUF → GGUF tensor transplant (135 of 144 targets); the 9 remaining expert tensors, which upstream stores as Q2_0, are re-encoded from the abliterated values using GSQ's own learned block scales — no GSQ value is recomputed, only the codes Not done Any dequantize → quantize cycle; any re-quantization from BF16 / safetensors In one line: what you get is still the GSQ-RCO quantization you already know , with those 144 tensors abliterated. It is the only abliteration route that preserves GSQ's work.
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
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。
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
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
🏆 BenchLocal Total: 4035/5000 (80.7%) — MTP Speculative Decoding Injected Uncensored · Abliterated · Agent-Optimized · 1.7-4.1× Speedup Uncensored version of QwenPaw-Flash-9B , processed with Heretic v1.3.0 abliteration, with MTP (Multi-Token Prediction) head weights injected from the original Qwen3.5-9B base model. By reconstructing the MTP speculative decoding head — which was stripped during the QwenPaw fine-tuning process — this model achieves up to 4.1× inference speedup on real agent benchmarks while maintaining or improving accuracy. Test Environment : NVIDIA RTX 5070 Ti (16GB) · llama.cpp (turboquant build, --spec-type draft-mtp ) · Q6_K quant Framework : BenchLocal — local model agent evaluation suite Methodology : Each scenario run once , no retries, no second attempts
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
FIT-GGUF v0.2.0 GATE-VERIFIED TIERS 4 TIERS · 2.26–3.15 GiB ABLITERATED (T615) MEASURED KL + SAME-TOP APACHE-2.0 Spark-X2.5-4B-abliterated · FIT-GGUF Four fidelity tiers of a 1M-context hybrid-attention 4B — every shipped file re-verified against its own BF16, byte-exact and reproducible. 🧭 About FIT-GGUF — the tool behind these files Every file in this repository was planned, executed and verified by FIT-GGUF , an open-source, deterministic tensor-level planning layer on top of llama.cpp quantization. Standard GGUF quantization asks you to pick one of a handful of presets; FIT-GGUF instead asks what quality do you want , then finds and verifies the smallest GGUF that demonstrably meets it. Traditional GGUF gives you presets. FIT gives you a fidelity contract: macro KL ≤ tier anchor ∧ same-top ≥ model-calibrated floor. Deterministic size prediction & byte-exact delivery ✅ Validated (G2 gate, delta = 0) Universally optimal tensor allocation ⚠️ Not established — FIT claims verified fidelity contracts, not a universal quality optimum The method, the preregistered research record and the fit CLI are open source: github.com/Scorp1o117/FIT-GGUF ⚠️ Safety notice / 安全提示 The source model is an abliterated, refusal-removed model with no meaningful built-in guardrails, and may comply with harmful, illegal or unsafe requests. Use it only where you can provide appropriate moderation, access control and legal review. Do not deploy it to end users without your own safety layer. 源模型经过拒答方向消融,不具备可靠的内置安全护栏;请仅在合法、受控、具备审核与滥用防护的环境中使用,使用者自行承担部署责任。
FIT-GGUF v0.5.0 ALL FIVE TIERS GATE-VERIFIED 5 FIT + BF16 + IMATRIX INPUT-GATED ABLITERATION MEASURED KL + SAME-TOP APACHE-2.0 Spark-X2.6-4B-abliterated · FIT-GGUF Five fidelity tiers of a 4B hybrid-attention model — each one the smallest GGUF that provably meets its KL gate , re-measured on its own shipped bytes. 🧭 About FIT-GGUF — the tool behind these files Every FIT tier here was planned, executed and verified by FIT-GGUF , an open-source, deterministic tensor-level planning layer on top of llama.cpp quantization. Standard GGUF quantization asks you to pick one of a handful of presets; FIT-GGUF instead asks what quality do you want , then finds and verifies the smallest GGUF that demonstrably meets it. Traditional GGUF gives you presets. FIT gives you a fidelity contract: macro KL ≤ tier anchor, measured against this model's own BF16 on five fixed domains. Deterministic size prediction & byte-exact delivery ✅ Validated (G2 gate, delta = 0) Universally optimal tensor allocation ⚠️ Not established — FIT claims verified fidelity contracts, not a universal quality optimum Method, pre-registered research record and the fit CLI are all open source: github.com/Scorp1o117/FIT-GGUF ⚠️ Safety notice / 安全提示 The source model is an abliterated, refusal-removed model with no meaningful built-in guardrails, and may comply with harmful, illegal or unsafe requests. Use it only where you can provide appropriate moderation, access control and legal review. Do not deploy it to end users without your own safety layer. 源模型经过拒答方向消融,不具备可靠的内置安全护栏;请仅在合法、受控、具备审核与滥用防护的环境中使用,使用者自行承担部署责任。
Swift 1.5 upstream · layout preserved 144 tensors transplanted IQ3_S / IQ3_XXS / IQ2_XS / Q2_0 per-tensor blake2b verified reasoning efficiency kept Swift Open License Swift-1.5-Qwen3.8-Flash-Next · GSQ-RCO-abliterated The refusal direction removed from UkisAI's reasoning-efficient derivative of Qwen3.8-Flash-Next. 144 tensors across all 48 layers swap in ready-made abliterated weights by byte-level transplant — every GSQ value and scale is left untouched, the file size moves by less than 0.2% , and the reasoning-efficiency of Swift 1.5 survives (still 34.8% below the base). 🧭 What this is The abliterated edition of UkisAI's Swift 1.5 — itself a reasoning-efficient derivative of Qwen3.8-Flash-Next that uses ~52% fewer thinking tokens on coding benchmarks. Every GSQ-learned quantized tensor is left bit-for-bit untouched; only the 144 tensors that write back into the residual stream are replaced, layer by layer, with the corresponding tensors from a ready-made abliterated release. Base ukisai/Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-GGUF —— GSQ learned quantization + RCO budget allocation, plus UkisAI’s anti-overthinking post-training Abliterated weights orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF —— the 144 target tensors’ bytes come straight from here Method GGUF → GGUF byte-level tensor transplant . Only where the official type demands it (the 48 routed-expert ffn_down_exps stored as Q2_0) are the abliterated values re-encoded — using Swift’s own learned block scales , so no GSQ value is recomputed, only the code words change. ⚠️ Safety notice This model is abliterated — the refusal direction has been removed and it has no reliable built-in guardrails . It may comply with harmful, illegal or unsafe requests. Deploy it only where you can supply your own moderation, access control and legal review; do not put it in front of end users without your own safety layer. Also note the upstream Swift Open License v1.0 is stricter than Apache-2.0: commercial use is free only below US$1M annual revenue.
No matching models
Adjust provider, category or input/output filters. New models are added continuously.
Great models. Clear costs.
Adjust monthly token usage to compare estimated API costs.
Compares base text-token rates within the same channel
Uses collected, comparable USD-per-million-token rates. Snapshots exclude cache discounts, tiers, extra reasoning output and taxes. Check current source prices.
Understand before you choose
How should API prices be compared?+
Separate input, output and cache rates, then estimate the same workload. Image, audio and video rates may use items, duration or resolution rather than tokens.
Do open weights mean free commercial use?+
Not necessarily. Check each version’s license, commercial restrictions and policies. Self-hosting also incurs compute and maintenance costs.
Is the top-ranked model the best for me?+
Rankings depend on tasks, versions and methods. Compare accuracy, latency, reliability and costs on your own workload.