IBM Granite 4.0
A Tier · 8.2/10
IBM's enterprise-focused open-weight family -- Granite 4.0 hybrid Mamba-2 + transformer architecture (70-80% memory reduction vs pure transformer), 3B to 32B sizes, Apache 2.0. First open model family to secure ISO 42001 certification. Nano 350M runs on CPU with 8-16GB RAM. 3B Vision variant landed 2026-04-01
Score Breakdown
The Good and the Bad
What we like
- +Hybrid Mamba-2 + transformer architecture delivers 70-80% memory reduction versus pure transformer at the same quality tier -- the practical result is that Granite 4.0 3B fits inside memory budgets that older 7B transformers needed, and Granite 32B fits where 70B would have
- +Granite 4.0 Nano (350M and 1.5B) is genuinely runnable on CPU with 8-16GB RAM -- no GPU required. One of the few open-weight options that realistically runs inside a browser via WebGPU or on a laptop without heat issues
- +First open model family with ISO 42001 AI management system certification -- matters significantly for regulated industries (healthcare, finance, government) where an audit trail on AI governance is a procurement requirement
- +Granite 4.0 3B Vision (2026-04-01) adds multimodality to the lineup at a size where consumer GPUs can actually run it. Apache 2.0 on the vision variant too
What could be better
- −Granite does not top absolute-quality open-weight benchmarks -- DeepSeek V3.2, Qwen 3.6, GLM-5.1, and Llama 4 all outscore it at the flagship tier. Granite competes on efficiency / governance / enterprise-fit, not on raw-model leaderboard position
- −Mamba-2 hybrid layers need first-party runtime support -- Ollama and llama.cpp support is improving but has been behind pure-transformer models. Check runtime compatibility before committing to Granite on exotic hardware
- −Smaller community than Llama / Qwen / DeepSeek -- fewer third-party fine-tunes, fewer Reddit help threads, less Stack Overflow coverage. IBM watsonx is the curated-docs route for teams that need that
- −IBM brand is polarizing in open-source -- some developers are enthusiastic about having a US enterprise incumbent in the open-weight game, others assume enterprise = stodgy and skip it entirely. Evaluate the models, not the logo
Pricing
Self-hosted (Apache 2.0)
- ✓Apache 2.0 license, unrestricted commercial use
- ✓Weights on Hugging Face + Ollama + watsonx
- ✓Full family: Nano 350M / 1.5B, 3B, 3B Vision, 7B, 32B
- ✓First open model family with ISO 42001 AI management system certification
watsonx.ai (IBM-hosted)
- ✓Enterprise-grade SLAs, data residency
- ✓ISO 42001 + SOC 2 + HIPAA compliance
- ✓Watsonx.governance integration (model cards, lineage)
- ✓Priced via IBM enterprise contracts
System Requirements
Hardware needed to self-host. Min = smallest viable setup (usually heavy quantization). Max = full-precision / production-grade.
| Model variant | Min | Max |
|---|---|---|
| Granite 4.0 Nano (350M / 1.5B)Apache 2.0. Runs on CPU with no GPU required | 8-16 GB RAM (CPU inference) | Any modern laptop GPU |
| Granite 4.0 3B / 3B Vision (2026-04-01)Mamba-2 hybrid; memory per token ~30 percent lower than pure transformer at same scale | 6 GB VRAM Q4 (RTX 3060) | 24 GB VRAM FP16 |
| Granite 4.0 7B / 32BISO 42001 certified, enterprise-grade governance | 8 / 24 GB VRAM Q4 | 1× A100 40 GB / 2× H100 FP16 |
Known Issues
- Granite 4.0 3B Vision released 2026-04-01 -- community tool support (llama.cpp vision tower, Ollama vision variant) is still catching up in the first 2-3 weeks post-launch. Watsonx.ai has first-class vision support out of the boxSource: IBM announcement, Hugging Face discussions · 2026-04
- Granite Nano 350M is a capability floor, not a ceiling -- it handles simple classification and short extraction well but struggles on multi-step reasoning. Use the 3B or larger for anything agenticSource: IBM model cards, VentureBeat coverage · 2025-12
Best for
Regulated-industry enterprises (healthcare, finance, government) who need Apache 2.0 open-weight models with ISO 42001 certification. Also ideal for edge deployments where Granite Nano (350M / 1.5B) is one of the few open models that runs realistically on CPU. And for any Mamba-hybrid research or low-memory production use where the 70-80% memory reduction actually changes the economics.
Not for
Teams chasing absolute raw quality -- DeepSeek V3.2, GLM-5.1, Qwen 3.6 are all stronger on general benchmarks. Also not for users who want a rich community ecosystem and extensive third-party fine-tunes -- Llama or Qwen wins on that axis.
Our Verdict
IBM Granite 4.0 is the best open-weight option in 2026 for enterprise procurement, edge deployment, and any workflow where the Mamba-2 hybrid architecture's memory savings materially change what hardware you need. The ISO 42001 certification is a genuinely differentiating feature for regulated industries -- nothing else in the open-weight category has it. The tradeoff is absolute-quality: Granite is not trying to beat DeepSeek or GLM on benchmarks, it is trying to be the model your compliance team can actually approve. If that matters, Granite is the pick. If it doesn't, you have other options that score higher.
Sources
- IBM: Granite 4.0 hyper-efficient hybrid models (accessed 2026-04-17)
- VentureBeat: Granite 4.0 Nano runs locally (accessed 2026-04-17)
- Hugging Face: IBM Granite 4.0 3B Vision (accessed 2026-04-17)
Explore more IBM Granite 4.0 rankings
Deeper leaderboards, benchmarks, task-specific tier lists, and status/pricing pages for IBM Granite 4.0.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
Alternatives to IBM Granite 4.0
Llama 4 (Meta)
Meta's open-weights family -- Scout (10M context), Maverick (multimodal 400B MoE). NOTE: Meta's frontier work moved to the proprietary Muse Spark line in April 2026; Llama remains downloadable and supported but is effectively in maintenance mode
Mistral AI
**New 2026-10-15: Timestral, Mistral's first time-series forecasting family (5M, 28M, 123M and 209M parameters, weights on Hugging Face since 10/14 under a non-Apache licence), with the 209M model claimed first on GIFT-Eval, TIME and fev-bench and CPU inference in 10-140 ms.** **Mistral Large 4 ('le Chonk', ML4) entered public preview on 2026-10-06: a 1T-parameter natively multimodal MoE with roughly 50B active parameters, a 1M context window, callable today as mistral-large-4 at $1.36 / $0.14 cached / $4.18 per 1M tokens (batch at half price), open weights promised by the end of October, and vendor-claimed open-weight leads in cyber (Cybench 93%), coding (DeepSWE v1.1 61.7%) and agentic work (AutomationBench 59.9%).** **Mistral opened a Munich hub for Physics AI and Industrial AI on 2026-09-28**, confirming the May 2026 acquisition of Emmi AI (30+ physicists and engineers in computational fluid dynamics, structural mechanics and multi-physics simulation), naming BMW (crash simulation) and Siemens Energy as partners, and committing to **build one gigawatt of European compute capacity by 2030**. **Mistral now powers Firefox Smart Window (2026-09-16)** -- Mozilla's beta AI browsing assistant runs on Mistral models for France and North America, UK and Germany later this year, with zero data retention. **EUR 3B Series D at a EUR 21B+ post-money (9/08, Samsung-led)**; Cloudera partnership (9/10); Medium 3 retired 8/31 in favour of Medium 3.5.
DeepSeek
**DeepSeek-V4.1-Flash-1017 (2026-10-17) replaced V4.1-Flash on the web, app and API at unchanged prices** (the deepseek-flash model name now serves it; MIT weights on Hugging Face), with higher scores and shorter outputs, and DeepSeek opened its in-house **Search and Fetch APIs at $0.75 per 1,000 calls per endpoint** -- server-side Search/Fetch calls through DeepSeek Harness, Claude Code and Codex stay free until 2026-10-31 16:00 UTC. Earlier: **DeepSeek-V4.1-Flash shipped 2026-09-10**: a 552B MoE on a new Causal Encoder-Decoder architecture (8B active for input, 16B for output), natively multimodal, MIT-licensed on HF, and priced BELOW the V4-Flash it replaces -- $0.15/$0.60 off-peak, $0.30/$1.20 peak per 1M, with cache-hit input cut to $0.003/$0.006. V4-Flash and V4-Flash-Vision-Exp are retired (aliases route to V4.1-Flash). DeepSeek said V4-Pro would be phased out on 9/14, then **reversed that four days later** -- V4-Pro stays on the API at unchanged $0.66/$1.98 off-peak, $1.32/$3.96 peak until V4.1-Pro lands. V4.1-Flash's vendor benchmarks beat V4-Pro (DeepSWE 74.2, Terminal-Bench 2.1 90.6, GPQA 90.9)
Gemma 4 (Google)
Google DeepMind's open-weights model family -- multimodal, 256K context, runs on edge devices **EmbeddingGemma 2 (announced 2026-10-06; weights on Hugging Face since 9/14) adds a 740M-parameter Apache 2.0 embedding model built on Gemma 4 that maps text, code, images, video and audio into one 768-dimensional space, with an 8K context, Matryoshka truncation down to 128 dimensions and about 191 MB of RAM for text-only use on a Pixel 11 Pro.**
DiffusionGemma (Google)
Google DeepMind's experimental open-weights TEXT-DIFFUSION model (June 10, 2026) -- 26B MoE (3.8B active), Apache 2.0, generates 256-token blocks in parallel with bidirectional attention for up to 4x faster output (1,000+ tok/s on H100). Trades some quality vs Gemma 4 for raw speed
Qwen (Alibaba)
**Qwen-Image-2.1 went open-weight (weights 2026-09-14, blog 2026-09-20)** -- a unified text-to-image and editing model with a 7B visual-generation component, native transparent (RGBA) output, up to 10 reference images and mask-guided local edits -- but under a **research-only licence, not Apache**: commercial use needs a separate licence from Alibaba. Alibaba's open-weights + API family, and August 2026 was its biggest open-release month yet: **Qwen3.8-2.4T-A95B -- the Max-class flagship -- went open-weight on 2026-08-08** (custom Qwen3.8-Max licence), **Qwen3.8-27B** dense landed Apache 2.0 (8/05), and **Qwen3.8-Flash-Next** (8/24 weights, 8/26 blog) previews the Qwen4 architecture: 125B main + 51B n-gram embeddings with only 6B active, served as Qwen3.8-Flash at $0.15/$0.47 per 1M. Qwen 3.7 Max remains the GA API flagship ($2.50/$7.50)
GLM / Z.ai (Zhipu AI)
**GLM 5.2 has a named production deployment at Perplexity: on 2026-09-21 Perplexity disclosed that Perplexity Computer sessions are served by GLM 5.2, post-trained by Perplexity on real user sessions** (21.2% relative cut in live tool-call failures). **GLM-5.3 and GLM-5.3-Flash released 2026-08-25** (weights on Hugging Face): 5.3 is a post-training-only upgrade of the 5.2 base that Z.ai calls 'the most capable open-weights model for coding' (Terminal Bench 2.1 88.2, DeepSWE 66.9, CyberGym 84.5) but moves from MIT to a custom licence; 5.3-Flash (320B/18B active, multimodal) stays MIT at $0.15/$0.50 per 1M.
Kimi K3 (Moonshot)
Moonshot's 2.8T-parameter Kimi K3 (launched 2026-07-16/17) is the largest open-weight model ever released -- 1M context, multimodal, $3/$15 per 1M via API, ranked best-available on Arena.AI at launch. WEIGHTS SHIPPED ~2026-07-26/27 on Hugging Face (2.8T total / 104B activated, safetensors) under a custom Kimi K3 License, not the Modified MIT of the K2 line
Nemotron (Nvidia)
Nvidia's open-weights family -- hybrid Mamba-Transformer MoE architecture, optimized for efficient reasoning on Nvidia hardware. Nemotron 3 Ultra (550B total / 55B active) shipped 2026-06-04 as the family flagship, joining Super (120B/12B, March) and Nano
MiniMax M3
**MiniMax-H3's weights are public on Hugging Face (repo 2026-07-28, updated 2026-08-13) under a community licence whose open-weight grant is limited to the US, EU, UK and South Korea**, and **MiniMax Music 3 (2026-08-07)** ships open weights for five-minute full-song generation with a prominent-attribution commercial clause. MiniMax's coding/agent flagship -- M3 (June 1 2026): 1M-token context, MSA sparse attention (>15x decoding speedup at long context), SWE-Bench Pro 59.0%, Terminal-Bench 66.0%. OPEN WEIGHTS LIVE on HuggingFace since June 12 (~428B total / ~23B active, native multimodal, minimax-community license)
Falcon (TII)
UAE's Technology Innovation Institute open-weights family -- Falcon 3 optimized for efficient sub-10B deployment on consumer hardware
gpt-oss (OpenAI)
OpenAI's FIRST open-weight models -- gpt-oss-120b (single 80GB GPU, near parity with o4-mini on reasoning) and gpt-oss-20b (runs on 16GB edge devices). Apache 2.0. Launched 2025-08-05. gpt-oss-safeguard ships in 2026 as the safety-tuned variant
Arcee Trinity-Large-Thinking
Arcee AI's US-made open-weight frontier reasoning model -- launched 2026-04-01. 398B total params, ~13B active. Sparse MoE (256 experts, 4 active = 1.56% routing). Apache 2.0, trained from scratch. #2 on PinchBench trailing only Claude 3.5 Opus. ~96% cheaper than Opus-4.6 on agentic tasks
Olmo 3 (AI2)
Allen Institute for AI's fully-open frontier reasoning models -- Olmo 3 family (2025-11-20) includes 7B and 32B sizes, four variants (Base, Think, Instruct, RLZero). Apache 2.0 with fully open data + checkpoints + training logs. Olmo 3-Think 32B matches Qwen3-32B-Thinking at 6x fewer training tokens
AI21 Jamba2
AI21 Labs' hybrid SSM-Transformer (Mamba-style) open-weight family -- Jamba2 launched 2026-01-08. Two sizes: 3B dense (runs on phones / laptops) and Jamba2 Mini MoE (12B active / 52B total). Apache 2.0, 256K context, mid-trained on 500B tokens
StepFun Step 3.7 Flash
StepFun's (China) agent-focused open-weight family -- Step 3.7 Flash (May 28 2026): 198B sparse MoE vision-language model, ~11B active, 256K context, Apache 2.0, ~400 tok/s, SWE-Bench Pro 56.3. Supersedes Step 3.5 Flash (Feb 2026) as the flagship
Cohere Command A
Cohere's enterprise-multilingual flagship -- 111B params, 256K context, runs on 2x H100. 23 languages. CC-BY-NC 4.0 on weights (research / non-commercial), commercial requires Cohere enterprise contract. Follow-ups: Command A Reasoning + Command A Vision
LongCat-2.0 (Meituan)
Meituan's open-source 1.6T-parameter MoE (~48B active) with native 1M-token context, MIT license -- trained entirely on domestic Chinese AI ASICs and revealed as the stealth 'Owl Alpha' model that had been topping OpenRouter
Inkling (Thinking Machines Lab)
Mira Murati's $12B lab ships its first model (2026-07-15): a 975B/41B-active open-weights MoE that reasons natively over text, images, and audio with a 1M-token context -- positioned not as the strongest model, but as the best starting point for fine-tuning via Tinker
Bonsai 27B (PrismML)
The first 27B-class model that runs on a phone (2026-07-14) -- ternary and 1-bit quantizations of Qwen3.6 27B squeeze a multimodal, tool-calling, 262K-context model into 3.9-5.9GB under Apache 2.0