will it local
Still the prosumer king. LocalScore 8B median: 120 tok/s gen, LocalScore 1727, TTFT 176 ms (accelerator/1704). TPU: 16384 CUDA / 512 Tensor cores, 21 Gbps effective GDDR6X, launched Sept 20 2022 at $1599.
computed band 90.0-150.0 tok/s for 8B Q4_K_M (roofline, 1008.0 GB/s VRAM); community: 120 tok/s Llama 3.1 8B Instruct Q4_K_M (https://www.localscore.ai/accelerator/1704)
Effective decode window: 0.46–0.76 of 1008 GB/s nominal → ~464–766 GB/s effective (llama.cpp decode, Q4_K_M basis; arch window, community-calibrated)
Community check — localmaxxing.com medians (1 bucket): localmaxxing.com
→ ~43 tok/s · 30-33B Q4 · 3 runs · just above our band (15% over the high edge)
→ 13b-q4 full 14.6 GB needed of 24 GB usable — headroom for context
→ 3b-q4 full 2.8 GB needed of 24 GB usable — headroom for context
→ 4b-q4 full 3.7 GB needed of 24 GB usable — headroom for context
→ 8b-q4 full 6.1 GB needed of 24 GB usable — headroom for context
→ 120 tok/s · Llama 3.1 8B Instruct Q4_K_M LocalScore median · source
Nominal bandwidth: 1008.0 GB/s — see the effective decode window above
→ qwen3:0.6b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 0.6B · 0.6GB · needs ~1GB (4k ctx)
→ qwen3:1.7b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 1.7B · 1.4GB · needs ~1.8GB (4k ctx)
→ qwen3:4b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 4B · 2.6GB · needs ~3.2GB (4k ctx)
→ llama3.2:1b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 1B · 1.3GB · needs ~1.4GB (4k ctx)
→ llama3.2:3b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 3B · 2.0GB · needs ~2.4GB (4k ctx)
→ gemma3:1b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 1B · 1.0GB · needs ~1.1GB (4k ctx)
→ phi4:mini on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) mini 3.8B · 2.5GB · needs ~3GB (4k ctx)
→ moondream:2b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 2B · 1.7GB · needs ~3.7GB (4k ctx)
→ whisper:tiny on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) tiny · 0.1GB · needs ~2.1GB (4k ctx)
→ whisper on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) base · 0.3GB · needs ~2.3GB (4k ctx)
→ whisper:small on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) small · 0.9GB · needs ~2.9GB (4k ctx)
→ whisper:medium on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) medium · 3.1GB · needs ~5.1GB (4k ctx)
→ kokoro:82m on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 82M · 0.3GB · needs ~2.3GB (4k ctx)
→ gemma3:4b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 4B · 3.3GB · needs ~3.8GB (4k ctx)
→ deepseek-r1:7b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) distill 7B · 4.7GB · needs ~4.9GB (4k ctx)
→ mistral:7b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 7B · 4.1GB · needs ~4.6GB (4k ctx)
→ llava:7b on Raspberry Pi 5 + AI Kit (Hailo-8L) ($110) 7B · 4.7GB · needs ~6.7GB (4k ctx)
→ qwen3:8b on Tesla P100 (16GB, used) ($135) 8B · 5.2GB · needs ~5.8GB (4k ctx)
→ qwen2.5vl:7b on Tesla P100 (16GB, used) ($135) 7B · 5.6GB · needs ~5.8GB (4k ctx)
→ llama3.1:8b on Tesla P100 (16GB, used) ($135) 8B · 4.9GB · needs ~5.4GB (4k ctx)
→ deepseek-r1:8b on Tesla P100 (16GB, used) ($135) distill 8B · 4.9GB · needs ~5.4GB (4k ctx)
→ whisper:large-v3 on Tesla P100 (16GB, used) ($135) large-v3 · 6.2GB · needs ~8.2GB (4k ctx)
→ mistral-nemo:12b on Tesla P100 (16GB, used) ($135) 12B · 7.1GB · needs ~7.7GB (4k ctx)
→ sdxl on Tesla P100 (16GB, used) ($135) SDXL base · 6.9GB · needs ~8.9GB (4k ctx)
→ llava:13b on Tesla P100 (16GB, used) ($135) 13B · 8.0GB · needs ~11.1GB (4k ctx)
→ gemma3:12b on Tesla P100 (16GB, used) ($135) 12B · 8.1GB · needs ~9.6GB (4k ctx)
→ qwen3:14b on Tesla P100 (16GB, used) ($135) 14B · 9.3GB · needs ~9.9GB (4k ctx)
→ deepseek-r1:14b on Tesla P100 (16GB, used) ($135) distill 14B · 9.0GB · needs ~9.8GB (4k ctx)
→ phi4:14b on Tesla P100 (16GB, used) ($135) 14B · 9.1GB · needs ~9.9GB (4k ctx)
→ gpt-oss:20b on Tesla P40 (24GB, used) ($290) 20B MoE · 14GB · needs ~14.2GB (4k ctx)
→ devstral:24b on Tesla P40 (24GB, used) ($290) 24B · 14.6GB · needs ~15.2GB (4k ctx)
→ qwen3.8:27b on Mac mini M4 (32GB) ($999) 27B · 16.5GB · needs ~18.5GB (4k ctx)
→ gemma3:27b on Mac mini M4 (32GB) ($999) 27B · 17GB · needs ~18.9GB (4k ctx)
→ flux:schnell on Mac mini M4 (32GB) ($999) schnell fp8 · 17GB · needs ~19GB (4k ctx)
→ qwen3:30b-a3b on Mac mini M4 (32GB) ($999) 30B-A3B MoE · 18.6GB · needs ~19GB (4k ctx)
→ qwen3-coder:30b-a3b on Mac mini M4 (32GB) ($999) 30B-A3B MoE · 18.6GB · needs ~19GB (4k ctx)
→ qwen3:32b on Mac mini M4 (32GB) ($999) 32B · 20GB · needs ~21GB (4k ctx)
→ deepseek-r1:32b on Mac mini M4 (32GB) ($999) distill 32B · 20GB · needs ~21GB (4k ctx)
→ qwen2.5vl:32b on Mac mini M5 Pro (64GB) ($1669) 32B · 22GB · needs ~23GB (4k ctx)
→ mixtral:8x7b 8x7B MoE — 26.5 GB vs 24 GB usable — partial CPU offload, expect large speed loss
→ flux:dev dev fp8 — 25.0 of 24 GB usable — barely over; real with q8 KV cache (halves KV) or shorter context
⚠ Used-market price trap: AI demand pushed used 4090s at/above the $1599 MSRP — a used card can cost more than a new 5080
⚠ 12V-2x6 connector melt incidents; inspect/reseat the power cable on any used unit
⚠ Production reportedly ended; a 4090 Ti / TITAN Ada never shipped, so 24GB is the Ada GeForce ceiling
⚠ Triple-slot ~450W card — verify case and PSU clearance before buying
⚠ LocalScore shows '23GB' on some listings — it's the same 24GB card (display quirk), not a different SKU
$5000 — 32GB GDDR7 / ~1.8 TB/s — the only meaningful same-form-factor step up for local LLM inference
$16499 — 48GB RTX Pro 6000 workstation card when 24GB is the binding constraint and used 4090 pricing is inflated anyway