Glyphae

State of the art.

The best AI models by category, in the cloud and on your own machine. Updated October 2026.

Text and code

ModelRunsBest forReleasedLicenseStatus
Claude Opus 5.5AnthropicCloudFable 5.1 level, 40% cheaper to run22 Sep 2026API
GPT-6 AstraOpenAICloudReasoning, computer use, coding4 Sep 2026API
Gemma 4 E2BGoogleLocalSmall model for text, images and audio. 3.2 GBIn Glyphae: Chat, spelling and titles in Glyphae AI2 Mar 2026Apache-2.0In Glyphae
Ternary Bonsai 2 27BPrismMLLocalA 27B model in ternary weights, 7.2 GB. GGUF, MLX or WebGPUIn Glyphae: Recaps and code in Glyphae AI16 Sep 2026Apache-2.0In Glyphae
DeepSeek V4.1 FlashDeepSeekCloudCheap structured analysis and JSON, under a cent a run10 Sep 2026API
GPT-6 Sol and LunaOpenAICloudGPT-6 at half the price22 Sep 2026API
Qwen3.8 27BQwenLocalThe most liked open model on Hugging Face5 Aug 2026Apache-2.0

Image

ModelRunsBest forReleasedLicenseStatus
GPT Image 2.5 SunburstOpenAICloudTop image arena score, up to 4K8 Sep 2026API
Qwen-Image 2.1QwenLocalGenerate, edit, transparent PNG. 4.2 GB14 Sep 2026ResearchTo test
Hy Image 3.5 previewTencentCloudUp to 4K, 20 reference images, 0.024 USD an image22 Sep 2026APITo test
QuiverAI Arrow 2QuiverAICloudEditable SVG: logos, icons, illustrations7 Sep 2026APITo test
Ming-Image 0.1 DesigninclusionAILocalPosters, UI, infographics with text17 Sep 2026MITTo test
Bonsai Image 4BPrismMLLocalFLUX.2 klein in 1 or 2 bits21 May 2026Apache-2.0To test
Z-Image-TurboTongyi-MAILocalFast, good for LoRAs25 Nov 2025Apache-2.0
Krea 2 TurboKreaLocalPhotographic, fast18 Jun 2026Community
Qwen-Image 2.1 TurboAlibabaLocalQwen-Image 2.1 in 8 steps, generate and edit. 7B9 Oct 2026Qwen Research LicenseTo test

Video

ModelRunsBest forReleasedLicenseStatus
Kling 3.0KuaishouCloudLeads blind votes. Native audio, 4K5 Feb 2026API
Wan2.2 5BWan-AILocalText and image to video, one GPU18 Jul 2025Apache-2.0
TesseractMirageLocalVideo editor your agent drives: cuts, motion, sound22 Sep 2026To test
LTX-2.5LightricksLocalMost downloaded open video model23 Jul 2026Community
MiniMax H3MiniMaxLocalVideo with stereo sound, 33B28 Jul 2026Excludes EU, US
Veo 3.1GoogleCloudStrong prompt adherence15 Oct 2025API
Seedance 2.5ByteDanceCloud30-second clips, local edits31 Jul 2026API
HunyuanVideo 1.5TencentLocalStrong image to video18 Nov 2025Excludes EU
Sora 2OpenAICloudAPI ends 24 September30 Sep 2025API

3D

ModelRunsBest forReleasedLicenseStatus
Tripo H3.1VASTCloudProduction meshes from an imageMar 2026API
TRELLIS.2MicrosoftLocalMesh with PBR textures1 Dec 2025MIT
Meshy 7.1MeshyCloudUltra 4K detail10 Sep 2026API
Hunyuan3D 3.5TencentCloud8K PBR textures, under a minuteAPI
Pixal3DTencentARCLocalOne image to GLB30 Apr 2026MIT

Music

ModelRunsBest forReleasedLicenseStatus
Suno v6SunoCloudFull songs with vocals9 Sep 2026API
ACE-Step 1.5 XLACE-StepLocalSongs with vocals, up to four minutes2 Apr 2026MIT
ElevenMusic v2.5ElevenLabsCloudSongs and scores14 Sep 2026API
MiniMax Music 3MiniMaxLocalSongs up to five minutes from lyrics7 Aug 2026Community
Stable Audio 3 MediumStability AILocalLoops, sound design, instrumentals17 May 2026Community
YuE2 3BM-A-PLocalSongs from lyrics9 Sep 2026To test
CLAPLAIONLocalAudio and text embeddings to search and compare soundsIn Glyphae: Finds similar tracks in Glyphae Music16 Feb 2023Apache-2.0In Glyphae

Voice

ModelRunsBest forReleasedLicenseStatus
Qwen3-TTS 1.7BQwenLocalSpeech in a custom voice21 Jan 2026Apache-2.0
Whisper large-v3 turboOpenAILocalSpeech to text, 99 languages1 Oct 2024MIT
VoxCPM2OpenBMBLocalVoice cloning from a short sample3 Apr 2026Apache-2.0
Kokoro 82MhexgradLocalTiny, fast text to speech26 Dec 2024Apache-2.0
Parakeet ReduxMoondreamLocal149M speech to text18 Sep 2026CC-BY-4.0

Decisions

ModelRunsBest forReleasedLicenseStatus
Jev 1.13TypeSafeCloudTyped answers with probabilities, no text15 Sep 2026API
Desert AntDesert Ant LabsLocalTiny on-device models: speech, PII, titles25 Aug 2026To test
Liquid d1-3BLiquid AILocalOpen-weight decisions from text and images. 6.3 GB7 Oct 2026LFM Open License 1.0
Clef-omniCloudflareCloudDecisions from text, image, audio and video in one call9 Oct 2026Apache-2.0
Clef and Clef-flashCloudflareCloudJev-compatible API, open weights. Text, images and video, 27B or 9B1 Oct 2026Apache-2.0
Kev-9BJared PalmerLocalJev-compatible API to run and fine-tune yourself. 6.4 GB20 Sep 2026Apache-2.0
TinyJev-4BAnkit AglaweLocalDecisions inside Ollama, reads documents up to 16k tokens. 4.5 GB1 Oct 2026Apache-2.0

Learn

NameKindWhy read it
SoL-PiLiu, Ye, Gao et al.PaperAn agent that improves its own harness: 45 to 49% fewer tokens
Building a harness with JevSydney Runkle, LangChainArticleSwap each decision in an agent loop for one typed question
GPT-6 Astra character workflowTripoArticleOne image to a rigged, animated character: Tripo makes the parts, Astra drives Blender
Tripo to a dancing characterNano, @Dstudio_aiArticleAn original character modeled, rigged and dancing in two days: Tripo, Blender, VRM 1.0

Build your own

ToolWhat it doesStatus
llama.cppggml-orgRun any open text model on your own machineIn Glyphae: Serves the local models of Glyphae AIIn Glyphae
stable-diffusion.cppleejetThe same for image modelsTo test
UnslothUnsloth AIFine-tune an open model on one GPUTo test
TRL 1.13Hugging FacePost-train a model: SFT, DPO, GRPOTo test
ai-toolkitOstrisTrain a LoRA: a style, a face, a productTo test