35 1 month ago

Flint1 is a dense 1.8B parameter language model exclusively trained by OakLeaf Labs. Despite its compact size, Flint1 delivers exceptional intelligence and reasoning capabilities, punching well above its weight class. Carefully crafted by OakLeaf.

thinking 1.8b
399faa591fc4 · 1.1GB
    Metadata
  • general.architecture
    qwen3
  • general.file_type
    Q4_K_M
  • qwen3.attention.head_count
    16
  • qwen3.attention.head_count_kv
    8
  • qwen3.attention.key_length
    128
  • qwen3.attention.layer_norm_rms_epsilon
    1e-06
  • qwen3.attention.value_length
    128
  • qwen3.block_count
    28
  • qwen3.context_length
    40960
  • qwen3.embedding_length
    2048
  • qwen3.feed_forward_length
    6144
  • qwen3.rope.freq_base
    1e+06
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.eos_token_id
    151645
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    151669
  • tokenizer.ggml.pre
    qwen2
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • Tensor
  • token_embd.weight
    Q6_K
    [2048, 151936]
  • blk.0
  • blk.0.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.0.attn_k_norm.weight
    F32
    [128]
  • blk.0.attn_norm.weight
    F32
    [2048]
  • blk.0.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.0.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.0.attn_q_norm.weight
    F32
    [128]
  • blk.0.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.0.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.0.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.0.ffn_norm.weight
    F32
    [2048]
  • blk.0.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.1
  • blk.1.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.1.attn_k_norm.weight
    F32
    [128]
  • blk.1.attn_norm.weight
    F32
    [2048]
  • blk.1.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.1.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.1.attn_q_norm.weight
    F32
    [128]
  • blk.1.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.1.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.1.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.1.ffn_norm.weight
    F32
    [2048]
  • blk.1.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.2
  • blk.2.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.2.attn_k_norm.weight
    F32
    [128]
  • blk.2.attn_norm.weight
    F32
    [2048]
  • blk.2.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.2.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.2.attn_q_norm.weight
    F32
    [128]
  • blk.2.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.2.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.2.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.2.ffn_norm.weight
    F32
    [2048]
  • blk.2.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.3
  • blk.3.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.3.attn_k_norm.weight
    F32
    [128]
  • blk.3.attn_norm.weight
    F32
    [2048]
  • blk.3.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.3.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.3.attn_q_norm.weight
    F32
    [128]
  • blk.3.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.3.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.3.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.3.ffn_norm.weight
    F32
    [2048]
  • blk.3.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.4
  • blk.4.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.4.attn_k_norm.weight
    F32
    [128]
  • blk.4.attn_norm.weight
    F32
    [2048]
  • blk.4.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.4.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.4.attn_q_norm.weight
    F32
    [128]
  • blk.4.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.4.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.4.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.4.ffn_norm.weight
    F32
    [2048]
  • blk.4.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.5
  • blk.5.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.5.attn_k_norm.weight
    F32
    [128]
  • blk.5.attn_norm.weight
    F32
    [2048]
  • blk.5.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.5.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.5.attn_q_norm.weight
    F32
    [128]
  • blk.5.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.5.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.5.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.5.ffn_norm.weight
    F32
    [2048]
  • blk.5.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.6
  • blk.6.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.6.attn_k_norm.weight
    F32
    [128]
  • blk.6.attn_norm.weight
    F32
    [2048]
  • blk.6.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.6.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.6.attn_q_norm.weight
    F32
    [128]
  • blk.6.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.6.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.6.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.6.ffn_norm.weight
    F32
    [2048]
  • blk.6.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.7
  • blk.7.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [128]
  • blk.7.attn_norm.weight
    F32
    [2048]
  • blk.7.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.7.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.7.attn_q_norm.weight
    F32
    [128]
  • blk.7.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.7.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.7.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.7.ffn_norm.weight
    F32
    [2048]
  • blk.7.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.8
  • blk.8.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.8.attn_k_norm.weight
    F32
    [128]
  • blk.8.attn_norm.weight
    F32
    [2048]
  • blk.8.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.8.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.8.attn_q_norm.weight
    F32
    [128]
  • blk.8.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.8.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.8.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.8.ffn_norm.weight
    F32
    [2048]
  • blk.8.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.9
  • blk.9.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.9.attn_k_norm.weight
    F32
    [128]
  • blk.9.attn_norm.weight
    F32
    [2048]
  • blk.9.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.9.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.9.attn_q_norm.weight
    F32
    [128]
  • blk.9.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.9.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.9.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.9.ffn_norm.weight
    F32
    [2048]
  • blk.9.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.10
  • blk.10.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.10.attn_k_norm.weight
    F32
    [128]
  • blk.10.attn_norm.weight
    F32
    [2048]
  • blk.10.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.10.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.10.attn_q_norm.weight
    F32
    [128]
  • blk.10.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.10.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.10.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.10.ffn_norm.weight
    F32
    [2048]
  • blk.10.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.11
  • blk.11.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.11.attn_k_norm.weight
    F32
    [128]
  • blk.11.attn_norm.weight
    F32
    [2048]
  • blk.11.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.11.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.11.attn_q_norm.weight
    F32
    [128]
  • blk.11.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.11.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.11.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.11.ffn_norm.weight
    F32
    [2048]
  • blk.11.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.12
  • blk.12.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.12.attn_k_norm.weight
    F32
    [128]
  • blk.12.attn_norm.weight
    F32
    [2048]
  • blk.12.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.12.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.12.attn_q_norm.weight
    F32
    [128]
  • blk.12.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.12.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.12.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.12.ffn_norm.weight
    F32
    [2048]
  • blk.12.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.13
  • blk.13.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.13.attn_k_norm.weight
    F32
    [128]
  • blk.13.attn_norm.weight
    F32
    [2048]
  • blk.13.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.13.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.13.attn_q_norm.weight
    F32
    [128]
  • blk.13.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.13.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.13.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.13.ffn_norm.weight
    F32
    [2048]
  • blk.13.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.14
  • blk.14.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.14.attn_k_norm.weight
    F32
    [128]
  • blk.14.attn_norm.weight
    F32
    [2048]
  • blk.14.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.14.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.14.attn_q_norm.weight
    F32
    [128]
  • blk.14.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.14.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.14.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.14.ffn_norm.weight
    F32
    [2048]
  • blk.14.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.15
  • blk.15.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.15.attn_k_norm.weight
    F32
    [128]
  • blk.15.attn_norm.weight
    F32
    [2048]
  • blk.15.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.15.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.15.attn_q_norm.weight
    F32
    [128]
  • blk.15.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.15.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.15.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.15.ffn_norm.weight
    F32
    [2048]
  • blk.15.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.16
  • blk.16.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.16.attn_k_norm.weight
    F32
    [128]
  • blk.16.attn_norm.weight
    F32
    [2048]
  • blk.16.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.16.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.16.attn_q_norm.weight
    F32
    [128]
  • blk.16.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.16.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.16.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.16.ffn_norm.weight
    F32
    [2048]
  • blk.16.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.17
  • blk.17.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.17.attn_k_norm.weight
    F32
    [128]
  • blk.17.attn_norm.weight
    F32
    [2048]
  • blk.17.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.17.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.17.attn_q_norm.weight
    F32
    [128]
  • blk.17.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.17.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.17.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.17.ffn_norm.weight
    F32
    [2048]
  • blk.17.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.18
  • blk.18.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.18.attn_k_norm.weight
    F32
    [128]
  • blk.18.attn_norm.weight
    F32
    [2048]
  • blk.18.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.18.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.18.attn_q_norm.weight
    F32
    [128]
  • blk.18.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.18.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.18.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.18.ffn_norm.weight
    F32
    [2048]
  • blk.18.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.19
  • blk.19.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.19.attn_k_norm.weight
    F32
    [128]
  • blk.19.attn_norm.weight
    F32
    [2048]
  • blk.19.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.19.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.19.attn_q_norm.weight
    F32
    [128]
  • blk.19.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.19.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.19.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.19.ffn_norm.weight
    F32
    [2048]
  • blk.19.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.20
  • blk.20.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.20.attn_k_norm.weight
    F32
    [128]
  • blk.20.attn_norm.weight
    F32
    [2048]
  • blk.20.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.20.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.20.attn_q_norm.weight
    F32
    [128]
  • blk.20.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.20.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.20.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.20.ffn_norm.weight
    F32
    [2048]
  • blk.20.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.21
  • blk.21.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.21.attn_k_norm.weight
    F32
    [128]
  • blk.21.attn_norm.weight
    F32
    [2048]
  • blk.21.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.21.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.21.attn_q_norm.weight
    F32
    [128]
  • blk.21.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.21.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.21.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.21.ffn_norm.weight
    F32
    [2048]
  • blk.21.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.22
  • blk.22.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.22.attn_k_norm.weight
    F32
    [128]
  • blk.22.attn_norm.weight
    F32
    [2048]
  • blk.22.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.22.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.22.attn_q_norm.weight
    F32
    [128]
  • blk.22.attn_v.weight
    Q4_K
    [2048, 1024]
  • blk.22.ffn_down.weight
    Q4_K
    [6144, 2048]
  • blk.22.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.22.ffn_norm.weight
    F32
    [2048]
  • blk.22.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.23
  • blk.23.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.23.attn_k_norm.weight
    F32
    [128]
  • blk.23.attn_norm.weight
    F32
    [2048]
  • blk.23.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.23.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.23.attn_q_norm.weight
    F32
    [128]
  • blk.23.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.23.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.23.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.23.ffn_norm.weight
    F32
    [2048]
  • blk.23.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.24
  • blk.24.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.24.attn_k_norm.weight
    F32
    [128]
  • blk.24.attn_norm.weight
    F32
    [2048]
  • blk.24.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.24.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.24.attn_q_norm.weight
    F32
    [128]
  • blk.24.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.24.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.24.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.24.ffn_norm.weight
    F32
    [2048]
  • blk.24.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.25
  • blk.25.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.25.attn_k_norm.weight
    F32
    [128]
  • blk.25.attn_norm.weight
    F32
    [2048]
  • blk.25.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.25.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.25.attn_q_norm.weight
    F32
    [128]
  • blk.25.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.25.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.25.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.25.ffn_norm.weight
    F32
    [2048]
  • blk.25.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.26
  • blk.26.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.26.attn_k_norm.weight
    F32
    [128]
  • blk.26.attn_norm.weight
    F32
    [2048]
  • blk.26.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.26.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.26.attn_q_norm.weight
    F32
    [128]
  • blk.26.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.26.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.26.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.26.ffn_norm.weight
    F32
    [2048]
  • blk.26.ffn_up.weight
    Q4_K
    [2048, 6144]
  • blk.27
  • blk.27.attn_k.weight
    Q4_K
    [2048, 1024]
  • blk.27.attn_k_norm.weight
    F32
    [128]
  • blk.27.attn_norm.weight
    F32
    [2048]
  • blk.27.attn_output.weight
    Q4_K
    [2048, 2048]
  • blk.27.attn_q.weight
    Q4_K
    [2048, 2048]
  • blk.27.attn_q_norm.weight
    F32
    [128]
  • blk.27.attn_v.weight
    Q6_K
    [2048, 1024]
  • blk.27.ffn_down.weight
    Q6_K
    [6144, 2048]
  • blk.27.ffn_gate.weight
    Q4_K
    [2048, 6144]
  • blk.27.ffn_norm.weight
    F32
    [2048]
  • blk.27.ffn_up.weight
    Q4_K
    [2048, 6144]
  • output_norm.weight
    F32
    [2048]