Qwen3.5 9B are supported?

[gabrieltotene]

Hello guys, i was trying to build my llamafile for Qwen3.5-9B-Q6_K_XL, but i get the follow problem when i run my builded llamafile:

Version: 0.10.0 of llamafile

`server_main: n_parallel is set to auto, using n_parallel = 4 and kv_unified = true
build: 1774356886 (7f5ee5496) with cosmocc for cosmopolitan
system info: n_threads = 6, n_threads_batch = 6, total_threads = 12

system_info: n_threads = 6 (n_threads_batch = 6) / 12 | CPU : LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 | CUDA : ARCHS = 750,800,860,890,900 | PEER_MAX_BATCH_SIZE = 128 |

init: using 11 threads for HTTP server
start: binding port with default address family
server_main: loading model
srv load_model: loading model '/zip/Qwen3.5-9B_Q6_K_XL.gguf'
common_init_result: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on
llama_params_fit_impl: projected to use 9004 MiB of device memory vs. 9272 MiB of free device memory
llama_params_fit_impl: cannot meet free memory target of 1024 MiB, need to reduce device memory by 755 MiB
llama_params_fit_impl: context size set by user to 32000 -> no change
llama_params_fit_impl: filling dense layers back-to-front:
llama_params_fit_impl: - CUDA0 (NVIDIA GeForce RTX 3060): 29 layers, 8086 MiB used, 1186 MiB free
llama_params_fit: successfully fit params to free device memory
llama_params_fit: fitting params to free memory took 2.55 seconds
llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 3060) (0000:09:00.0) - 9474 MiB free
llama_model_loader: loaded meta data with 46 key-value pairs and 427 tensors from /zip/Qwen3.5-9B_Q6_K_XL.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = qwen35
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Qwen3.5-9B
llama_model_loader: - kv 3: general.basename str = Qwen3.5-9B
llama_model_loader: - kv 4: general.quantized_by str = Unsloth
llama_model_loader: - kv 5: general.size_label str = 9B
llama_model_loader: - kv 6: general.license str = apache-2.0
llama_model_loader: - kv 7: general.license.link str = https://huggingface.co/Qwen/Qwen3.5-9...
llama_model_loader: - kv 8: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 9: general.base_model.count u32 = 1
llama_model_loader: - kv 10: general.base_model.0.name str = Qwen3.5 9B
llama_model_loader: - kv 11: general.base_model.0.organization str = Qwen
llama_model_loader: - kv 12: general.base_model.0.repo_url str = https://huggingface.co/Qwen/Qwen3.5-9B
llama_model_loader: - kv 13: general.tags arr[str,2] = ["unsloth", "image-text-to-text"]
llama_model_loader: - kv 14: qwen35.block_count u32 = 32
llama_model_loader: - kv 15: qwen35.context_length u32 = 262144
llama_model_loader: - kv 16: qwen35.embedding_length u32 = 4096
llama_model_loader: - kv 17: qwen35.feed_forward_length u32 = 12288
llama_model_loader: - kv 18: qwen35.attention.head_count u32 = 16
llama_model_loader: - kv 19: qwen35.attention.head_count_kv u32 = 4
llama_model_loader: - kv 20: qwen35.rope.dimension_sections arr[i32,4] = [11, 11, 10, 0]
llama_model_loader: - kv 21: qwen35.rope.freq_base f32 = 10000000.000000
llama_model_loader: - kv 22: qwen35.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 23: qwen35.attention.key_length u32 = 256
llama_model_loader: - kv 24: qwen35.attention.value_length u32 = 256
llama_model_loader: - kv 25: qwen35.ssm.conv_kernel u32 = 4
llama_model_loader: - kv 26: qwen35.ssm.state_size u32 = 128
llama_model_loader: - kv 27: qwen35.ssm.group_count u32 = 16
llama_model_loader: - kv 28: qwen35.ssm.time_step_rank u32 = 32
llama_model_loader: - kv 29: qwen35.ssm.inner_size u32 = 4096
llama_model_loader: - kv 30: qwen35.full_attention_interval u32 = 4
llama_model_loader: - kv 31: qwen35.rope.dimension_count u32 = 64
llama_model_loader: - kv 32: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 33: tokenizer.ggml.pre str = qwen35
llama_model_loader: - kv 34: tokenizer.ggml.tokens arr[str,248320] = ["!", """, "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 35: tokenizer.ggml.token_type arr[i32,248320] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 36: tokenizer.ggml.merges arr[str,247587] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv 37: tokenizer.ggml.eos_token_id u32 = 248046
llama_model_loader: - kv 38: tokenizer.ggml.padding_token_id u32 = 248055
llama_model_loader: - kv 39: tokenizer.chat_template str = {%- set image_count = namespace(value...
llama_model_loader: - kv 40: general.quantization_version u32 = 2
llama_model_loader: - kv 41: general.file_type u32 = 18
llama_model_loader: - kv 42: quantize.imatrix.file str = Qwen3.5-9B-GGUF/imatrix_unsloth.gguf
llama_model_loader: - kv 43: quantize.imatrix.dataset str = unsloth_calibration_Qwen3.5-9B.txt
llama_model_loader: - kv 44: quantize.imatrix.entries_count u32 = 248
llama_model_loader: - kv 45: quantize.imatrix.chunks_count u32 = 80
llama_model_loader: - type f32: 177 tensors
llama_model_loader: - type f16: 72 tensors
llama_model_loader: - type q8_0: 70 tensors
llama_model_loader: - type q6_K: 108 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q6_K
print_info: file size = 8.15 GiB (7.81 BPW)
load: 0 unused tokens
load: printing all EOG tokens:
load: - 248044 ('<|endoftext|>')
load: - 248046 ('<|im_end|>')
load: - 248063 ('<|fim_pad|>')
load: - 248064 ('<|repo_name|>')
load: - 248065 ('<|file_sep|>')
load: special tokens cache size = 33
load: token to piece cache size = 1.7581 MB
print_info: arch = qwen35
print_info: vocab_only = 0
print_info: no_alloc = 0
print_info: n_ctx_train = 262144
print_info: n_embd = 4096
print_info: n_embd_inp = 4096
print_info: n_layer = 32
print_info: n_head = 16
print_info: n_head_kv = 4
print_info: n_rot = 64
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 256
print_info: n_embd_head_v = 256
print_info: n_gqa = 4
print_info: n_embd_k_gqa = 1024
print_info: n_embd_v_gqa = 1024
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-06
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 0.0e+00
print_info: n_ff = 12288
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: n_expert_groups = 0
print_info: n_group_used = 0
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 40
print_info: rope scaling = linear
print_info: freq_base_train = 10000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 262144
print_info: rope_yarn_log_mul = 0.0000
print_info: rope_finetuned = unknown
print_info: mrope sections = [11, 11, 10, 0]
print_info: ssm_d_conv = 4
print_info: ssm_d_inner = 4096
print_info: ssm_d_state = 128
print_info: ssm_dt_rank = 32
print_info: ssm_n_group = 16
print_info: ssm_dt_b_c_rms = 0
print_info: model type = ?B
print_info: model params = 8.95 B
print_info: general.name = Qwen3.5-9B
print_info: vocab type = BPE
print_info: n_vocab = 248320
print_info: n_merges = 247587
print_info: BOS token = 11 ','
print_info: EOS token = 248046 '<|im_end|>'
print_info: EOT token = 248046 '<|im_end|>'
print_info: PAD token = 248055 '<|vision_pad|>'
print_info: LF token = 198 'Ċ'
print_info: FIM PRE token = 248060 '<|fim_prefix|>'
print_info: FIM SUF token = 248062 '<|fim_suffix|>'
print_info: FIM MID token = 248061 '<|fim_middle|>'
print_info: FIM PAD token = 248063 '<|fim_pad|>'
print_info: FIM REP token = 248064 '<|repo_name|>'
print_info: FIM SEP token = 248065 '<|file_sep|>'
print_info: EOG token = 248044 '<|endoftext|>'
print_info: EOG token = 248046 '<|im_end|>'
print_info: EOG token = 248063 '<|fim_pad|>'
print_info: EOG token = 248064 '<|repo_name|>'
print_info: EOG token = 248065 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true, direct_io = false)
load_tensors: offloading output layer to GPU
load_tensors: offloading 28 repeating layers to GPU
load_tensors: offloaded 29/33 layers to GPU
load_tensors: CPU_Mapped model buffer size = 1798.13 MiB
load_tensors: CUDA0 model buffer size = 6542.67 MiB
..............................................................................
common_init_result: added <|endoftext|> logit bias = -inf
common_init_result: added <|im_end|> logit bias = -inf
common_init_result: added <|fim_pad|> logit bias = -inf
common_init_result: added <|repo_name|> logit bias = -inf
common_init_result: added <|file_sep|> logit bias = -inf
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 32000
llama_context: n_ctx_seq = 32000
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = auto
llama_context: kv_unified = true
llama_context: freq_base = 10000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_seq (32000) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
llama_context: CUDA_Host output buffer size = 3.79 MiB
llama_kv_cache: CPU KV buffer size = 125.00 MiB
llama_kv_cache: CUDA0 KV buffer size = 875.00 MiB
llama_kv_cache: size = 1000.00 MiB ( 32000 cells, 8 layers, 4/1 seqs), K (f16): 500.00 MiB, V (f16): 500.00 MiB
llama_memory_recurrent: CPU RS buffer size = 25.12 MiB
llama_memory_recurrent: CUDA0 RS buffer size = 175.88 MiB
llama_memory_recurrent: size = 201.00 MiB ( 4 cells, 32 layers, 4 seqs), R (f32): 9.00 MiB, S (f32): 192.00 MiB
sched_reserve: reserving ...
sched_reserve: Flash Attention was auto, set to enabled
sched_reserve: CUDA0 compute buffer size = 493.00 MiB
sched_reserve: CUDA_Host compute buffer size = 102.50 MiB
sched_reserve: graph nodes = 4209 (with bs=512), 2361 (with bs=1)
sched_reserve: graph splits = 309 (with bs=512), 10 (with bs=1)
sched_reserve: reserve took 84.43 ms, sched copies = 1
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
/home/gabriel.totene/Documentos/github-branchs/llamafile/llamafile/../llama.cpp/ggml/src/ggml-cuda/solve_tri.cu:279: solve_tri with n > 64 or k > 32 requires cuBLAS TRSM which is not available with TinyBLAS. This operation is only used by Qwen3-Next models. Please rebuild with cuBLAS or use CPU backend for this model.
[New LWP 3048687]
[New LWP 3048686]
[New LWP 3048685]
[New LWP 3048684]
[New LWP 3048683]
[New LWP 3048613]
[New LWP 3048612]
[New LWP 3048611]
[New LWP 3048610]
[New LWP 3048609]
[New LWP 3048608]
[New LWP 3048607]
[New LWP 3048606]
[New LWP 3048605]
[New LWP 3048604]
[New LWP 3048603]
[New LWP 3048602]
[New LWP 3048601]
[New LWP 3048600]
[New LWP 3048599]
[New LWP 3048598]
[New LWP 3048593]
aviso: A handler for the OS ABI "FreeBSD" is not built into this configuration
of GDB. Attempting to continue with the default i386:x86-64 settings.

aviso: A handler for the OS ABI "FreeBSD" is not built into this configuration
of GDB. Attempting to continue with the default i386:x86-64 settings.

aviso: Architecture rejected target-supplied description
Recursive internal problem.
Abortado (imagem do núcleo gravada)
`
I've tried to build the llamafile with ggml-cuda-cublas.so but i get the same error, any Ideas?

[aittalam]

Hi @gabrieltotene! The problem has been solved from 0.10.1 on (it contains this PR which reimplements the missing function in tinyblas). Actually, I just released 0.10.2 so I'd suggest you to try that out instead ;-)

Both the llamafile binary you find in the 0.10.2 release page and any of the prebuilt llamafiles (including some models from the Qwen3.5 family) now come with pre-bundled tinyblas cuda libs.

Also, it looks like you are running from FreeBSD, which is an OS I haven't had a chance to test the new llamafiles on yet. Please feel free to reach out if you are having any issues, I want to make sure they run on BSDs too 🙏