Home / 32 GB

32 GB · 2× RTX 5060 Ti 16 GB (32 GB in total)

Hy4-Flash 220B on two RTX 5060 Ti: Project Maya v1.1.68 with the Maya-XS quant (setup defaults)

Weights: Maya-XS GGUF quant, 44.8 GB (made by Peasant Smith, peasantsmith/Hy4-Flash-Maya-GGUF, of tencent/Hy4-flash)

2x NVIDIA GeForce RTX 5060 Ti 16 GB (two GPUs, layer split). Host: Intel Core i7-12700K (20 threads), 31 GiB RAM, NVMe.

Engine and setup (Project Maya v1.1.68, commit 05e0e2da) and the Maya quants of Hy4-FlashPeasant Smith (mw00)X @PeasantSmith
Project Maya maya-hy4 v1.1.68 (mw00/project-maya @05e0e2da), local build with user CUDA 13.3 toolkit, sm_120, MTP drafting oncommit mayaproject 05e0e2da2 GPU (layer split)measured 2026-11-20

Not ranked yet. The rank needs a finished, verified Gauntlet race (the game test), and this setup has not played it yet. It shows Speed v2, Quality v2 and the memory test, every number with its raw files and exact command. How the game test works.

Speed v2

How fast it runs

Decode
6.7tok/s
Decode, code
6.8tok/s
Decode, prose
6.5tok/s
500 words
≈ 103.0s
Prefill at 8K
124.5tok/s
First token at 2K
21.2s
MeasurementResult
Decode6.7 tok/s
Words per second4.9 words/s
Characters per second32.2 chars/s
Decode, code prompt6.8 tok/s
Decode, prose prompt6.5 tok/s
Spread of the decode runs (CV)18.7 %
Spread of the code runs (CV)18.7 %
Spread of the prose runs (CV)6.8 %
Decode, window 16.2 tok/s
Prefill at 2K99.9 tok/s
Prefill at 8K124.5 tok/s
Time to first token at 2K21.2 s
Time to first token at 8K68.3 s
Draft acceptance rate62.3 %
1 request at once (total)5.3 tok/s
2 at once (total)5.4 tok/s
4 at once (total)5.5 tok/s
8 at once (total)5.2 tok/s
1 request at once (per request)5.3 tok/s
2 at once (per request)4.0 tok/s
4 at once (per request)2.3 tok/s
8 at once (per request)1.2 tok/s
Server slots (the engine answers one request at a time; the setup offers no setting)1 slot

Long prompts

Prompt lengthPrefillTime to first tokenDecode, code promptDecode, prose prompt
16K tokens140.4 tok/s116.7 s6.4 tok/s6.7 tok/s

Method: code+prose; 2 run(s) each; 512 tokens; 1 warm-up(s); median; context steps: prompts of exactly 16K, 32K, 64K and 128K tokens (a filler log first, the code or prose task prompt last), 2 runs per step, 512 output tokens; decode = tokens after the first / time after the first token; prefill = prompt tokens / time to first token of the same request, median over its requests; requests at once: C1,C2,C4,C8 at once; 2 rounds each; 512 tokens per request; prose prompt; total = all tokens / wall time, per request = tokens / end-to-end time; median over the rounds. Measured 2026-11-20. Raw files: speed result · speed contract · log · long answer.

What ran. Project Maya v1.1.68 (mw00/project-maya, tag v1.1.68, commit 05e0e2da90cacf174d791b0192f3488a3c61100b, the release as the author ships it: cloned and installed with the project's own setup script), Tencent's Hy4-Flash (220B parameters in total, 12B active) in the Maya quant Maya-XS (44.8 GB), with the drafting block (MTP) switched on (--mtp as written by the setup) as the setup chooses it. The experts of the model are split between the card and the CPU: the engine's start line reads 'CPU lane: 4469 of 20992 routed experts in VRAM (21%), the rest on 12 CPU threads from the model's file'. Maya-XS (44.8 GB, 44,786,173,984 bytes), the smaller Maya quant of Hy4-Flash; Tencent's model card lists the Maya-XS at 88.5 % and the Maya-L at 96.6 % 'same top token as the original', so the two quants are not the same file and the numbers of the pages of the RTX 5090 are not the numbers of this quant. The engine did not take tensor parallel: the start line says a member would need 24.32 GB of VRAM and the first card has less free, so it runs the layer split over the two cards (30.0 GB of VRAM over 2 GPUs). The setup's defaults are used: window 32,768 tokens, prompt chunk chosen by the engine (it chose 4,096 tokens), drafting on.

The headline 6.7 tok/s is the mean of the code prompt (6.8) and the prose prompt (6.5), medians of two runs each (spread (CV) 18.7 % and 6.8 %). The code prompt has 1,220 and 1,221 tokens and the prose prompt 1,305 and 1,305 tokens; every reply is 512 tokens long and ends at the limit. About 103.0 s for 500 words (4.86 words per second).

Spread. The code prompt has a spread (CV) of 18.7 % between its two runs, higher than on the other Hy4 pages; the two values of the code prompt are 5.9 tok/s and 7.7 tok/s. Accepted drafts: 62.3 % (1,156 of 1,857).

Long prompts. The server window is 32,768 tokens: the prefill series has the sizes 2K and 8K (99.9 and 124.5 tok/s, first token after 21.2 s and 68.3 s) and the 16K context step ran (prefill 140.4 tok/s, first token after 116.7 s); the steps at 32K, 64K and 128K do not fit the window and were not sent (rows speed2_ctx_d32768_fits, speed2_ctx_d65536_fits and speed2_ctx_d131072_fits are 0).

Requests at once. Hy4-Flash answers one request at a time in Project Maya (docs/SETTINGS.md line 2385), so the levels 2, 4 and 8 at once are above its one slot: the requests wait in a queue and run one after another, not in parallel. The total stays near the value of one request (5.3, 5.4, 5.5 and 5.2 tok/s with 1, 2, 4 and 8 requests) and the speed of each request falls as the queue grows (5.3, 4.0, 2.3 and 1.2 tok/s).

Sampling. The speed test sent the sampling of the model card, which is the same for instruct and thinking (temperature 0.9, top_p 1.0), with the flag for no thinking (enable_thinking false), seed 42 and cache_prompt false; Tencent's card names reasoning_effort high (thinking) or no_think, and the test uses no_think. Cache check: the check of the test tool is marked not ok. 41 of the 46 requests that it checked report cached tokens: 6 or 7 tokens on the decode and requests-at-once requests and 22 or 23 tokens on the prefill requests. The Maya server keeps the first tokens of the chat template (the template head) in its cache on every request, so a fully cold prompt is not possible with this server; each request still carries a nonce, so no earlier answer or long prompt body is reused. The result is shown as measured.

Foreign load. The load guard of the speed test repeated 1 of 26 steps (1 repeats) (details in the guard section of the speed result).

Variant. The page of the same setup with a prompt chunk of 1,024 tokens and drafting off (the 'chunk 1024' page) is the faster one on these cards; this page is the setup's default.

Exact command we measured withfor reproducing↓

You do not need this to run the model. It is the exact command of our measurement, so anyone can repeat the result.

Speed v2 (run hy4-maya-xs-2011c-speed-r0)
python3 speed2.py --base-url http://127.0.0.1:33449/v1 --label hy4-maya-xs-2011c-speed-r0 --prompts $BENCH/repos/rigmark-v2/prompts-v2.json --output $RUN/raw/hy4-maya-xs-2011c-speed-r0/hy4-maya-xs-2011c-speed-r0/speed2 --contract $RUN/raw/hy4-maya-xs-2011c-speed-r0/hy4-maya-xs-2011c-speed-r0/speed2/speed-contract.json --parts decode,prefill,conc,l1,ctx --prefill-depths 2048,8192,32768 --prompt-ids code,prose --runs 2 --max-tokens 512 --warmups 1 --prefill-runs 3 --conc-levels 1,2,4,8 --conc-rounds 2 --conc-max-tokens 512 --conc-prompt prose --ctx-depths 16384,32768,65536,131072 --ctx-runs 2 --ctx-max-tokens 512 --sampling {"temperature": 0.9, "top_p": 1.0} --sampling-source Tencent Hy4-flash generation_config.json as applied by Project Maya (maya_quants.MODEL_DEFAULTS['hy4']; docs/SETTINGS.md:2343-2352: temperature 0.9, top-p 1.0, no top-k) and the model card peasantsmith/Hy4-Flash-Maya-GGUF 'Recommended Settings'. Greedy is discouraged by the author (hf-modelcard). --model hy4-flash --no-cache-prompt --server-log $RUN/logs/server-A.log --engine-version-file $RUN/logs/engine-version.txt --server-slots 1 --server-slots-source docs/SETTINGS.md:2385 (Hy4-Flash: --parallel not available, one request at a time) and serve/server.py:7 (one sequence at a time behind a FIFO) --server-ctx 32768 --server-ctx-source card box-5-hy4-flash-maya-xs-mayahy4-v1168-2x5060ti-2011 field ctx
Quality v2

Quality

Not run on this setup: no Quality v2 test, so the page shows no quality number.

Credits

Also made this possible

  • NVIDIA engine for Hy4 (maya-hy4, contributed to Project Maya in pull requests #1040 and #1051)needmorevramGitHub
  • Model (Hy4-Flash, 220B parameters in total, 12B active)TencentHugging Face
  • Quants (Maya-L and Maya-XS GGUF files)peasantsmithHugging Face
Setup

Run it yourself

What you need and how to start it, step by step. The results below come from exactly this setup.

What you need

Hardware
2x NVIDIA GeForce RTX 5060 Ti 16 GB (two GPUs, layer split). Host: Intel Core i7-12700K (20 threads), 31 GiB RAM, NVMe.
Model
peasantsmith/Hy4-Flash-Maya-GGUF Maya-XS (Project Maya's own quant of tencent/Hy4-flash, 220B, sha256 equal to the pinned table of v1.1.68)
Model (Hy4-Flash, 220B parameters in total, 12B active)
huggingface.co/tencent/Hy4-flash
Quants (Maya-L and Maya-XS GGUF files)
huggingface.co/peasantsmith/Hy4-Flash-Maya-GGUF
Engine
Project Maya as shipped (setup defaults, drafting on); most of the model runs on the CPU; Speed v2 only — the program that loads the model and answers requests · github.com/mw00/project-mayaBuild and settings we used: Project Maya maya-hy4 v1.1.68 (mw00/project-maya @05e0e2da), local build with user CUDA 13.3 toolkit, sm_120, MTP drafting on

Steps

  1. Get the engine and the model files from the list above.

  2. Start the server with this command. It loads the model into memory.

    bash <start script of the run>
Start text for this setupfor reproducing↓

You do not need this to run the model. It is the start text of this setup, based on the setup card of this run, with local paths, machine names and addresses shortened or replaced by placeholders, notes added on some pages, and the engine build in front, so anyone can repeat the result.

Start, based on the setup card of this run
Project Maya maya-hy4 v1.1.68 (mw00/project-maya @05e0e2da), local build with user CUDA 13.3 toolkit, sm_120, MTP drafting on; bash <start script of the run> (starts: .venv/bin/python maya.py --config maya-hy4-flash-maya-xs.json --port <port> --plain, with CUDA_VISIBLE_DEVICES set to the two cards; config written by ./maya.sh --setup --model hy4-flash-maya-xs; engine args as written by the setup: --pack <Maya-XS folder> --max-context 32768 --mtp)

Context and sampling: 32,768 tokens (the setup's recommendation for these cards), prompt chunk chosen by the engine (auto, it chose 4,096 tokens), drafting on (--mtp), vision on, one request at a time. Speed test with thinking off per request, sampling of the model card (temperature 0.9, top_p 1.0), seed 42.. Hardware: 2x NVIDIA GeForce RTX 5060 Ti 16 GB (two GPUs, layer split). Host: Intel Core i7-12700K (20 threads), 31 GiB RAM, NVMe.. All raw files of this setup: raw/v2/hy4-flash-2x5060ti-maya-xs-project-maya-1168-mw00.