v0.1.49-beta

Hey guys we got lots of new update for Unsloth, especially customization. Unsloth now yours to personalize: three color palettes plus custom colors and fonts, seven new display languages, and a new Voice settings tab for dictation and read-aloud. Agents get safer with a four-level tool-call permission selector (Ask, Approve for me, Off, Full access) and workspace isolation, and Intel GPUs finally get GPU-accelerated inference through new Vulkan `llama.cpp` support.

 [Inkling](https://unsloth.ai/docs/models/inkling), a new 975B parameter (41B active) open model with up to a 1M context window. Licensed under Apache 2.0, Inkling accepts text, images, and audio and generates text. Unsloth Studio natively supports it!

## Personalize Your Studio
Appearance is now fully customizable, not just light/dark (#7077):
- **Three color palettes**, each with its own light and dark scheme: **Standard** (the Unsloth green), **Classic** (a neutral enterprise grey/black/white with a sparing blue accent), and **Minimal** (strictly monochrome).
- **Custom accent, background, and foreground colors** through an in-app color picker (hue/saturation, hex, eyedropper), and **custom UI, heading, chat, and code fonts** with font import (woff2/woff/ttf/otf) and device/folder font detection.
- Font-size, contrast, pointer-cursor, reduced-motion, and font-smoothing controls, plus a restyle of the core controls (inputs, dropdowns, buttons) and settings search. Every setting is per-mode (light and dark keep their own values) and syncs across devices.

### Seven New Languages
- Added French, German, Spanish, Hindi, Arabic, Russian, and Korean display languages (#7076), joining the existing Chinese (Simplified), Japanese, and Portuguese (Brazil), and made Auto-detect (from the browser language) the new default.

### Voice Settings
- New Voice settings tab for dictation, dictionary, and read-aloud (#7074). (Full voice conversation mode is still in progress and not part of this release.)
- STT and TTS coming very soon!

## Tool-Call Permissions & Safer Agents
Finer control and safer defaults whenever a model calls a tool or an agent runs.
- The Bypass-permissions toggle becomes a four-level tool-call permission selector: **Ask** (approve every call), **Approve for me** (auto-runs read-only calls, pauses only on potentially-unsafe ones), **Off**, and **Full access** (also disables the sandbox). Covers terminal, python, web search, RAG, and MCP tools (#7079).
- Agent workspace isolation and Hermes one-shot resume (#7103); `unsloth start` now warns before running an agent's remote installer (#7024) and gains `--persist` to keep and reopen agent sessions (#7014).
- Live tool output streams with SSE heartbeats, with better web-page extraction and surfaced interrupted turns (#7083); binary `web_search` fetches are rejected instead of decoded into replacement characters (#7130).
- Removed `AGENTS.md` / `CLAUDE.md` from install artifacts and tightened instruction-file cleanup (#7096, #7097); the seeded bootstrap password auto-fill is now scoped to loopback clients (#7131).

## Dynamic NVFP4
See our Dynamic NVFP4 article for more details: https://unsloth.ai/docs/basics/nvfp4
We also shipped a broader **[NVFP4 collection](https://huggingface.co/collections/unsloth/nvfp4)** this release:
* [Qwen3.6 35B](https://huggingface.co/unsloth/Qwen3.6-35B-A3B-NVFP4) and (Qwen3.6 35B [Fast](https://huggingface.co/unsloth/Qwen3.6-35B-A3B-NVFP4-Fast)) and [27B](https://huggingface.co/unsloth/Qwen3.6-27B-NVFP4)
* [Qwen3.5 122B](https://huggingface.co/unsloth/Qwen3.5-122B-A10B-NVFP4)
* [Inkling](https://huggingface.co/unsloth/inkling-NVFP4)
* [GLM-4.7 Flash](https://huggingface.co/unsloth/GLM-4.7-Flash-NVFP4)
* Gemma 4: [31B](https://huggingface.co/unsloth/gemma-4-31B-it-NVFP4), [26B](https://huggingface.co/unsloth/gemma-4-26B-A4B-it-NVFP4), [12B](https://huggingface.co/unsloth/gemma-4-12b-it-NVFP4), [E4B](https://huggingface.co/unsloth/gemma-4-E4B-it-NVFP4), [E2B](https://huggingface.co/unsloth/gemma-4-E2B-it-NVFP4)

## Vulkan & llama.cpp
- New Vulkan `llama.cpp` backend: Intel GPUs now get GPU-accelerated inference instead of falling back to CPU (they use Vulkan, not CUDA/ROCm), reusing the existing VRAM, context auto-sizing, multi-GPU selection, and layer-offload paths. AMD users can opt in with `UNSLOTH_FORCE_VULKAN=1` (#5819).
- Prebuilts resolve via the release-assets CDN to avoid GitHub API rate limits (#7086); update-apply is pinned to the exact release the banner offered (#7112); fixed the permanent GGUF "update available" on no-symlink caches (#7113); model state resyncs after a `llama.cpp` update unloads it (#6998).
- Stop now interrupts a `llama.cpp` generation stalled mid-stream (#7117).

## Models & Training
- DeepSeek-V4: eager attention and trainable FP8 grouped experts (#7042).
- Stopped silent full-sequence training: "Train on completions" used a hardcoded template table and silently disabled masking on a miss, so DeepSeek and most LFM models trained on prompts too while the toggle said otherwise. Markers are now auto-detected from the tokenizer's chat template (#7054); also fixed manual response-template markers that never matched their rendered templates (#7062).
- Honor the "none" gradient-checkpointing option instead of silently forcing "unsloth" (#7128).
- Restore the process-global `torch.compile` config on torch 2.12 so gradient-checkpointing backward honors it (#7019).
- Keep native RoPE scaling when extending context, carrying `rope_theta` for linear scaling (#7028).
- Force-terminate a stuck training stop after a grace period (#7099).
- Route new architectures automatically: `lfm2_moe` (LFM2-8B-A1B) to transformers 5.3.0 (#7040), models by `CONFIG_MAPPING_NAMES` instead of hardcoded tables (#7043), and offer the latest transformers release for brand-new architectures (#7056).
- Inkling (TML) support fixes for multi-GPU B200s: KV-cache-aware context buffer sizing, tool-call healing for its typed `<|content_invoke_tool_json|>` markers, a download-worker `HF_TOKEN` fallback, and numeric reasoning-effort mapping (#7153).

## FP8, Quant & Export
- Fixed fast inference on compressed-tensors FP8 models (#7025), restored dropped FP8 `weight_scale_inv` tensors on load (#6978), and guarded FP8 Triton launches with tensor device context (#6888).
- Single-pass GGUF export for directly convertible outtypes in `save.py` (#7090); concise NVFP4 inference errors (#7145).
- Probe xformers support on `sm_120` instead of disabling it by version (#6828).

## OpenAI-Compatible API, GGUF & MCP
- Hardened OpenAI-compatible GGUF streaming (#6950) and queue local GGUF requests before `llama-server` (#7047).
- `/v1/messages` accepts thinking and null/unknown content blocks, fixing resumed Claude Code / Anthropic sessions that previously 400'd (#7017); the "No model loaded" 400 now hints at Model auto-switch (#7006); Studio auto-loads the last used local model, including GGUF variant (#6966).
- Render image content returned by MCP tools (#7081); persistent stdio MCP sessions keep server state across tool calls (#7080).

## Recipes, RAG & Data Prep
- Fixed `SyntheticDataKit.chunk_data` emitting chunks over `max_tokens` (#7073) and guarded `smart_chunk_text` against `stride >= chunk_size` (#7126).
- Fixed `FastSentenceTransformer` Qwen embedding preprocessing (#6939) and switched recipes to a writable artifact path (#7044).

## Installer, Hardware & Platform
- Fixed the Windows installer torch index override (#6972) and install a torchao Windows ROCm stub in the inference worker (#7000).
- Added the `UNSLOTH_SKIP_AUTOSTART` installer flag (#7093); the installer now returns proper failure exit codes (#7123).
- Detect Windows Intel GPUs via the registry before WMI (#7064) and expose Windows drive roots in the folder browser (#7082).

## Polish & Reliability
- Fixed Studio auto-titles for reasoning models (#7098), user-message overflow for long unbroken text (#7100), and duplicate response model labels and hover (#7049).
- Recover MLX VLM image prompts (#7094); prevent an auth-monitor reload loop (#7118); fix S3 tab flashing on reload (#7106) and Hub offline status (#7129).
- Quiet noisy logs, log real progress, and speed up Windows/macOS dataset prep (#7087); added a startup loading banner and muted the benign bitsandbytes ROCm warning (#7085).
- UI polish: compact thinking control in narrow composers (#7150), tidier permission composer layout and Hub feed icons (#7148), toast close-button positioning (#7142), and removed the edge-fades appearance setting (#7143).
- Made the Cloudflare tunnel opt-in, off by default (#7046).


# Updating Unsloth
To update Unsloth or install a fresh Unsloth Unsloth, use the commands below:
**macOS, Linux, WSL:**
```
curl -fsSL https://unsloth.ai/install.sh | sh
```
**Windows:**
```
irm https://unsloth.ai/install.ps1 | iex
```

## What's Changed
* Bump install.sh / install.ps1 pin to unsloth>=2026.7.1 by @danielhanchen in https://github.com/unslothai/unsloth/pull/6943
* Sort chat recents by last activity by @NilayYadav in https://github.com/unslothai/unsloth/pull/6844
* Studio: render \[ \] and \( \) LaTeX delimiters in chat by @oobabooga in https://github.com/unslothai/unsloth/pull/6914
* fix: match qwen3-thinking double-newline in train_on_responses_only response pattern by @InfoSage05 in https://github.com/unslothai/unsloth/pull/6926
* Studio: stream reasoning tokens in the tool-loop generator (fixes DeepSeek thinking not streaming with a pill on) by @oobabooga in https://github.com/unslothai/unsloth/pull/6947
* Create ossf.yml by @danielhanchen in https://github.com/unslothai/unsloth/pull/6952
* Speed up Studio startup path by @wasimysaid in https://github.com/unslothai/unsloth/pull/6899
* Polish assistant message actions menu by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6962
* Move New badge to System settings tab by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6963
* Fix opencode Unsloth provider selection by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6906
* Fix Hermes install hint on Windows by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6903
* Studio: heal DiffusionGemma tool calls into structured tool_calls by @oobabooga in https://github.com/unslothai/unsloth/pull/6851
* Fix case-variant model matching and GGUF cache reuse in unsloth start by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6900
* Studio: show Hugging Face address on hover for Hub and online model rows (#6382) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6928
* Studio: fix currency and indentation edge cases in LaTeX rendering by @danielhanchen in https://github.com/unslothai/unsloth/pull/6957
* Add MLX backend support for CLI unsloth train by @Lyxot in https://github.com/unslothai/unsloth/pull/6709
* feat(cli): support MLX distributed inference by @Lyxot in https://github.com/unslothai/unsloth/pull/6845
* Route MLX trainer callbacks through UnslothTrainer by @Lyxot in https://github.com/unslothai/unsloth/pull/6929
* (GRPO) Fix PEFT replacement for TRL >= 1.7.0, add missing compute_aux_loss for TRL >= 1.7.0 by @marcandrelarochelle in https://github.com/unslothai/unsloth/pull/6904
* version-compat CI: fake CPU training runs for SFT/GRPO/DPO by @danielhanchen in https://github.com/unslothai/unsloth/pull/6965
* Fix OpenClaw start default to local TUI by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6937
* feat: detect installed coding agent CLIs in Studio settings by @ErenAta16 in https://github.com/unslothai/unsloth/pull/6909
* Studio: don't pin transformers before the training worker activates the 5.x sidecar by @danielhanchen in https://github.com/unslothai/unsloth/pull/6968
* Studio: source CPU llama.cpp prebuilts from unslothai/llama.cpp by @oobabooga in https://github.com/unslothai/unsloth/pull/6311
* fix(studio/hub): apply repo_id length limit per segment, not whole string (#6946) by @Anai-Guo in https://github.com/unslothai/unsloth/pull/6953
* MoE LoRA: auto-target per-expert Linear experts (gpt-oss 4bit) instead of leaving them frozen by @danielhanchen in https://github.com/unslothai/unsloth/pull/6936
* Studio: fix flash-attn and torchao install on Blackwell (sm_100+) GPUs (Closes #6961) by @ThomasEricB in https://github.com/unslothai/unsloth/pull/6970
* Fix Backend CI: add has_blackwell_gpu to the mlx worker test stub by @danielhanchen in https://github.com/unslothai/unsloth/pull/6980
* Studio: allow CPU-only DiffusionGemma by granting the diffusion runner the CPU device by @danielhanchen in https://github.com/unslothai/unsloth/pull/6979
* Bump install.sh / install.ps1 pin to unsloth>=2026.7.2 by @danielhanchen in https://github.com/unslothai/unsloth/pull/6981
* Studio: render thinking blocks for safetensors inference with prefilled <think> templates by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6816
* Remove API menu new badge by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6983
* Fix BAD_MAPPINGS not redirecting the -unsloth-bnb-4bit dynamic quants by @vineethsaivs in https://github.com/unslothai/unsloth/pull/6949
* Fix duplicate gemma-2b-bnb-4bit key routing the base model to the instruct model by @vineethsaivs in https://github.com/unslothai/unsloth/pull/6891
* Fix to_sharegpt optional block rendering "None" for missing extra columns by @vineethsaivs in https://github.com/unslothai/unsloth/pull/6827
* Guard FP8 Triton launches with tensor device context by @ramisworld in https://github.com/unslothai/unsloth/pull/6888
* Fix per-block ID collisions and add block cleanup for unstructured uploads by @NilayYadav in https://github.com/unslothai/unsloth/pull/6944
* Stabilize floating monitor drag by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6984
* Retry the Studio UI shutdown re-login on transient goto timeout by @danielhanchen in https://github.com/unslothai/unsloth/pull/7027
* Fix FastSentenceTransformer Qwen embedding preprocessing by @Etherll in https://github.com/unslothai/unsloth/pull/6939
* unsloth start: warn before running an agent's remote installer by @danielhanchen in https://github.com/unslothai/unsloth/pull/7024
* Restore process-global torch.compile config on torch 2.12 so gradient checkpointing backward honors it by @danielhanchen in https://github.com/unslothai/unsloth/pull/7019
* Silence torch._check_is_size FutureWarning and shim it if torch removes it by @danielhanchen in https://github.com/unslothai/unsloth/pull/7023
* unsloth start: add --persist to keep and reopen agent sessions by @danielhanchen in https://github.com/unslothai/unsloth/pull/7014
* Studio /v1/messages: accept thinking and unknown content blocks by @danielhanchen in https://github.com/unslothai/unsloth/pull/7017
* Studio: add Vulkan llama.cpp support by @oobabooga in https://github.com/unslothai/unsloth/pull/5819
* Fix Windows installer torch index override by @alkinun in https://github.com/unslothai/unsloth/pull/6972
* Fix fast inference crash on compressed-tensors FP8 models by @danielhanchen in https://github.com/unslothai/unsloth/pull/7025
* Keep native RoPE scaling when extending context; carry rope_theta for linear by @danielhanchen in https://github.com/unslothai/unsloth/pull/7028
* scripts: refresh scan_packages allowlist baseline by @danielhanchen in https://github.com/unslothai/unsloth/pull/7032
* Studio: remove dead direct_linux_release_plan path by @danielhanchen in https://github.com/unslothai/unsloth/pull/7030
* Restore dropped FP8 weight_scale_inv tensors on load by @danielhanchen in https://github.com/unslothai/unsloth/pull/6978
* Studio: resolve the repo-root MTP drafter after the MTP/ GGUF rename by @danielhanchen in https://github.com/unslothai/unsloth/pull/7031
* Studio: harden OpenAI-compatible GGUF streaming by @Apoze in https://github.com/unslothai/unsloth/pull/6950
* Studio: auto-load last used local model by @alkinun in https://github.com/unslothai/unsloth/pull/6966
* Clarify in README that -H 0.0.0.0 starts a public Cloudflare tunnel by @oobabooga in https://github.com/unslothai/unsloth/pull/7007
* CI: retry transient HTTP timeouts in Studio smoke probes by @danielhanchen in https://github.com/unslothai/unsloth/pull/7052
* fix: guard remove_special_tokens against tokenizers without a BOS token by @vineethsaivs in https://github.com/unslothai/unsloth/pull/7048
* Studio: queue local GGUF OpenAI-compatible requests before llama-server by @Apoze in https://github.com/unslothai/unsloth/pull/7047
* Studio: hint at Model auto-switch in the OpenAI "No model loaded" 400 by @oobabooga in https://github.com/unslothai/unsloth/pull/7006
* Studio: detect Windows Intel GPUs via the registry before WMI by @oobabooga in https://github.com/unslothai/unsloth/pull/7064
* Studio: route lfm2_moe (LFM2-8B-A1B) to transformers 5.3.0 by @danielhanchen in https://github.com/unslothai/unsloth/pull/7040
* Studio: route models by CONFIG_MAPPING_NAMES instead of hardcoded tables by @danielhanchen in https://github.com/unslothai/unsloth/pull/7043
* Studio: auto-detect completion masking markers, stop silent full-sequence training by @danielhanchen in https://github.com/unslothai/unsloth/pull/7054
* scripts: refresh scan_packages allowlist baseline by @danielhanchen in https://github.com/unslothai/unsloth/pull/7078
* DeepSeek-V4: eager attention and trainable FP8 grouped experts by @danielhanchen in https://github.com/unslothai/unsloth/pull/7042
* Studio: fix the manual response-template markers that never match their rendered templates by @danielhanchen in https://github.com/unslothai/unsloth/pull/7062
* Fix SyntheticDataKit.chunk_data emitting chunks over max_tokens by @winklemad in https://github.com/unslothai/unsloth/pull/7073
* Studio: startup loading banner and mute the benign bitsandbytes ROCm warning by @danielhanchen in https://github.com/unslothai/unsloth/pull/7085
* Studio: add UNSLOTH_SKIP_AUTOSTART installer flag by @danielhanchen in https://github.com/unslothai/unsloth/pull/7093
* Studio: remove AGENTS.md and CLAUDE.md from install artifacts by @danielhanchen in https://github.com/unslothai/unsloth/pull/7096
* Tighten Studio instruction-file cleanup boundaries by @danielhanchen in https://github.com/unslothai/unsloth/pull/7097
* Fix Studio user-message overflow for long unbroken text by @Lyxot in https://github.com/unslothai/unsloth/pull/7100
* fix(studio): use writable recipe artifact path by @Lyxot in https://github.com/unslothai/unsloth/pull/7044
* Fix Studio auto-titles for reasoning models by @Lyxot in https://github.com/unslothai/unsloth/pull/7098
* Studio: resync model state after a llama.cpp update unloads it by @oobabooga in https://github.com/unslothai/unsloth/pull/6998
* Studio: Fix torch_dtype deprecation warning on startup and ASR load by @oobabooga in https://github.com/unslothai/unsloth/pull/6999
* Studio: Show Run button for downloaded non-GGUF models in the Model Hub by @oobabooga in https://github.com/unslothai/unsloth/pull/7001
* Studio: install torchao Windows ROCm stub in the inference worker by @oobabooga in https://github.com/unslothai/unsloth/pull/7000
* Studio: Add rename to project chat rows by @oobabooga in https://github.com/unslothai/unsloth/pull/7005
* Probe xformers support on sm_120 instead of disabling it by version by @oobabooga in https://github.com/unslothai/unsloth/pull/6828
* Studio: render image content returned by MCP tools by @NilayYadav in https://github.com/unslothai/unsloth/pull/7081
* Studio: add French, German, Spanish, Hindi, Arabic, Russian and Korean display languages by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7076
* scan_packages: baseline sentencepiece dup2 finding after upstream reindent by @danielhanchen in https://github.com/unslothai/unsloth/pull/7120
* Studio: persistent stdio MCP sessions so server state survives across tool calls by @NilayYadav in https://github.com/unslothai/unsloth/pull/7080
* Studio: resolve llama.cpp prebuilts via the release-assets CDN to avoid GitHub API rate limits by @danielhanchen in https://github.com/unslothai/unsloth/pull/7086
* fix(studio): prevent auth monitor reload loop by @Lyxot in https://github.com/unslothai/unsloth/pull/7118
* Studio: make Stop interrupt a llama.cpp generation stalled mid-stream by @oobabooga in https://github.com/unslothai/unsloth/pull/7117
* Unsloth: appearance palettes, customization options, and control restyle by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7077
* Studio: pin llama.cpp update apply to the release the banner offered by @oobabooga in https://github.com/unslothai/unsloth/pull/7112
* Studio: fix the permanent GGUF "update available" on no-symlink caches by @gaurav0107 in https://github.com/unslothai/unsloth/pull/7113
* Fix S3 tab flashing on reload by @NilayYadav in https://github.com/unslothai/unsloth/pull/7106
* fix: correct DeepScaleR-1.5B 16bit mapping to its own repo by @anxkhn in https://github.com/unslothai/unsloth/pull/7088
* fix(dataprep): guard smart_chunk_text against stride >= chunk_size by @anxkhn in https://github.com/unslothai/unsloth/pull/7126
* Fix Hub offline status by @NilayYadav in https://github.com/unslothai/unsloth/pull/7129
* fix: name unsloth_vllm_standby parameter in vLLM standby error by @anxkhn in https://github.com/unslothai/unsloth/pull/7089
* Studio: expose Windows drive roots in the folder browser by @gaurav0107 in https://github.com/unslothai/unsloth/pull/7082
* Studio CI: make tool-calling SSE probes resilient to transport stalls by @danielhanchen in https://github.com/unslothai/unsloth/pull/7137
* Studio: force-terminate a stuck training stop after a grace period by @danielhanchen in https://github.com/unslothai/unsloth/pull/7099
* fix(studio): recover MLX VLM image prompts by @Lyxot in https://github.com/unslothai/unsloth/pull/7094
* fix(install): return failure exit codes by @ShiroKSH in https://github.com/unslothai/unsloth/pull/7123
* Fix bare except clauses and remove duplicate MAX_FUSED_SIZE definition by @lxcxjxhx in https://github.com/unslothai/unsloth/pull/7138
* Studio: scope the seeded bootstrap password auto-fill to loopback clients by @oobabooga in https://github.com/unslothai/unsloth/pull/7131
* fix: single-pass GGUF export for directly convertible outtypes in save.py by @dylanschroers in https://github.com/unslothai/unsloth/pull/7090
* Studio: offer the latest transformers release for brand-new architectures by @danielhanchen in https://github.com/unslothai/unsloth/pull/7056
* Fix agent workspace isolation and Hermes one-shot resume by @wasimysaid in https://github.com/unslothai/unsloth/pull/7103
* Studio: permission levels for chat tool calls (Ask, Approve for me, Off, Full access) by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7079
* Studio: make the Cloudflare tunnel opt-in (off by default) by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/7046
* Studio: honor the 'none' gradient checkpointing option in training by @oobabooga in https://github.com/unslothai/unsloth/pull/7128
* Studio: remove the edge fades appearance setting by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7143
* Studio: quiet noisy logs, log real progress, and speed up Windows/macOS dataset prep by @danielhanchen in https://github.com/unslothai/unsloth/pull/7087
* Show concise NVFP4 inference errors by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7145
* Fix Studio toast close-button positioning by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7142
* Studio: add Voice settings tab (dictation, dictionary, read aloud) by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7074
* Studio: stream live tool output with SSE heartbeats, fix web page extraction, and surface interrupted turns by @danielhanchen in https://github.com/unslothai/unsloth/pull/7083
* Studio: fix permission composer layout and Hub feed icons by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7148
* Compact thinking control in narrow composers by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/7150
* Studio: reject binary web_search fetches instead of decoding them into replacement chars by @oobabooga in https://github.com/unslothai/unsloth/pull/7130
* Studio: fix duplicate response model labels and hover by @Etherll in https://github.com/unslothai/unsloth/pull/7049

## New Contributors
* @ErenAta16 made their first contribution in https://github.com/unslothai/unsloth/pull/6909
* @ThomasEricB made their first contribution in https://github.com/unslothai/unsloth/pull/6970
* @Apoze made their first contribution in https://github.com/unslothai/unsloth/pull/6950
* @winklemad made their first contribution in https://github.com/unslothai/unsloth/pull/7073
* @gaurav0107 made their first contribution in https://github.com/unslothai/unsloth/pull/7113
* @ShiroKSH made their first contribution in https://github.com/unslothai/unsloth/pull/7123
* @lxcxjxhx made their first contribution in https://github.com/unslothai/unsloth/pull/7138

**Full Changelog**: https://github.com/unslothai/unsloth/compare/v0.1.48-beta...v0.1.49-beta