CI for Spanish tests: limit the models

The focus is to prioritize:
* the smallest variant of the same model
* 32K context length or more
* best cost for token generation

Author: ariya

.github/workflows/spanish.yml

@@ -12,29 +12,13 @@ jobs:
       fail-fast: false
       matrix:
         model:
-          - meta-llama/llama-3.2-3b-instruct             # $0.0100/$0.0200 [ 128K]
-          - mistralai/mistral-nemo                       # $0.0100/$0.0270 [ 128K]
           - google/gemma-3n-e4b-it                       # $0.0200/$0.0400 [  32K]
           - google/gemma-3-4b-it                         # $0.0200/$0.0400 [ 128K]
           - mistralai/ministral-3b                       # $0.0400/$0.0400 [ 128K]
-          - mistralai/mistral-7b-instruct-v0.3           # $0.0280/$0.0540 [  32K]
           - mistralai/mistral-small-3.2-24b-instruct     # $0.0500/$0.1000 [  32K]
           - google/gemma-3-12b-it                        # $0.0500/$0.1000 [ 128K]
-          - microsoft/phi-3-mini-128k-instruct           # $0.1000/$0.1000 [ 128K]
           - qwen/qwen3-8b                                # $0.0350/$0.1380 [ 128K]
-          - amazon/nova-micro-v1                         # $0.0350/$0.1400 [ 128K]
-          - microsoft/phi-4                              # $0.0700/$0.1400 [  16K]
           - google/gemini-flash-1.5-8b                   # $0.0380/$0.1500 [1000K]
-          - amazon/nova-lite-v1                          # $0.0600/$0.2400 [ 300K]
-          - qwen/qwen3-14b                               # $0.0800/$0.2400 [  40K]
-          - mistralai/mixtral-8x7b-instruct              # $0.2400/$0.2400 [  32K]
-          - meta-llama/llama-3.3-70b-instruct            # $0.0700/$0.2500 [ 128K]
-          - qwen/qwen3-30b-a3b                           # $0.0800/$0.2900 [  40K]
-          - google/gemini-2.0-flash-lite-001             # $0.0750/$0.3000 [1000K]
-          - meta-llama/llama-4-scout                     # $0.0800/$0.3000 [ 320K]
-          - google/gemini-2.0-flash-001                  # $0.1000/$0.4000 [1000K]
-          - openai/gpt-4.1-nano                          # $0.1000/$0.4000 [1000K]
-          - google/gemini-2.5-flash-lite-preview-06-17   # $0.1000/$0.4000 [1000K]
 
     steps:
       - uses: actions/checkout@v4
@@ -60,29 +44,13 @@ jobs:
       fail-fast: false
       matrix:
         model:
-          - meta-llama/llama-3.2-3b-instruct             # $0.0100/$0.0200 [ 128K]
-          - mistralai/mistral-nemo                       # $0.0100/$0.0270 [ 128K]
           - google/gemma-3n-e4b-it                       # $0.0200/$0.0400 [  32K]
           - google/gemma-3-4b-it                         # $0.0200/$0.0400 [ 128K]
           - mistralai/ministral-3b                       # $0.0400/$0.0400 [ 128K]
-          - mistralai/mistral-7b-instruct-v0.3           # $0.0280/$0.0540 [  32K]
           - mistralai/mistral-small-3.2-24b-instruct     # $0.0500/$0.1000 [  32K]
           - google/gemma-3-12b-it                        # $0.0500/$0.1000 [ 128K]
-          - microsoft/phi-3-mini-128k-instruct           # $0.1000/$0.1000 [ 128K]          
           - qwen/qwen3-8b                                # $0.0350/$0.1380 [ 128K]
-          - amazon/nova-micro-v1                         # $0.0350/$0.1400 [ 128K]
-          - microsoft/phi-4                              # $0.0700/$0.1400 [  16K]
           - google/gemini-flash-1.5-8b                   # $0.0380/$0.1500 [1000K]
-          - amazon/nova-lite-v1                          # $0.0600/$0.2400 [ 300K]
-          - qwen/qwen3-14b                               # $0.0800/$0.2400 [  40K]
-          - mistralai/mixtral-8x7b-instruct              # $0.2400/$0.2400 [  32K]
-          - meta-llama/llama-3.3-70b-instruct            # $0.0700/$0.2500 [ 128K]
-          - qwen/qwen3-30b-a3b                           # $0.0800/$0.2900 [  40K]
-          - google/gemini-2.0-flash-lite-001             # $0.0750/$0.3000 [1000K]
-          - meta-llama/llama-4-scout                     # $0.0800/$0.3000 [ 320K]
-          - google/gemini-2.0-flash-001                  # $0.1000/$0.4000 [1000K]
-          - openai/gpt-4.1-nano                          # $0.1000/$0.4000 [1000K]
-          - google/gemini-2.5-flash-lite-preview-06-17   # $0.1000/$0.4000 [1000K]
 
     steps:
       - uses: actions/checkout@v4