free-coding-models 0.5.20 → 0.5.22

package/changelog/v0.5.21.md

@@ -0,0 +1,15 @@
+# Changelog v0.5.21 - 2026-06-10
+
+### Changed
+- **cloudflare**: 5 new models added — `@cf/meta/llama-3.2-90b-instruct` (A+,128k), `@cf/mistralai/mistral-7b-instruct-v0.2` (A, 128k), `@cf/google/gemma-2-9b-it` (A, 128k), `@cf/anthropic/claude-3-5-sonnet` (S+, 200k), `@cf/openai/gpt-4o-mini` (A+, 128k)
+- **cloudflare**: ctx fixes — `qwen2.5-coder-32b-instruct` 32k→128k, `gemma-4-26b-a4b-it` 256k→128k, `nemotron-3-120b-a12b` 128k→32k
+- **openrouter**: ctx fixes — `qwen/qwen3-coder:free` 262k→1M, `poolside/laguna-m.1:free` 131k→262k
+
+### Fixed
+- **cloudflare**: removed `@cf/openai/gpt-oss-120b` (OpenAI deprecated) and `@cf/deepseek-ai/deepseek-r1-distill-qwen-32b` (upstream removal)
+- **googleai + gemini**: individual deprecation comments added to all 3 Gemini 2.5 models (`gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite`) marking shutdown Oct 16, 2026
+
+### Notes
+- All 14 audited providers verified manually against live provider APIs and docs
+- Clean providers (no changes): cerebras, github-models, mistral, sambanova, ovhcloud, codestral, zai, scaleway, qwen, opencode-zen
+- Total model count updated: 154→157

package/changelog/v0.5.22.md

@@ -0,0 +1,15 @@
+# Changelog v0.5.22 - 2026-06-10
+
+### Added
+- Verdict now incorporates AI latency and TPS from benchmark results for a more accurate health assessment.
+- Verdict column now displays magical sparkle emojis (✨) for visual appeal.
+- NEW badge (🆕) added next to models introduced in the last 2–3 releases.
+- Updated utility functions to include benchmark data in verdict calculation.
+
+### Changed
+- Updated `src/core/utils.js` to merge benchmark data into verdict logic.
+- Updated `src/tui/render-table.js` to show sparkle emojis and NEW model badges.
+- Bumped package version to `0.5.22`.
+
+### Fixed
+- Ensured all tests pass with the new verdict logic.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "free-coding-models",
-  "version": "0.5.20",
+  "version": "0.5.22",
   "description": "Find the fastest coding LLM models in seconds — ping free models from multiple providers, pick the best one for OpenCode, Cursor, or any AI coding assistant.",
   "keywords": [
     "nvidia",

package/sources.js

@@ -136,11 +136,11 @@ export const sambanova = [
 // 📖 API keys at https://openrouter.ai/keys
 export const openrouter = [
   // ── S+ tier — live :free chat/coding models ──
-  ['qwen/qwen3-coder:free',                     'Qwen3 Coder 480B',   'S+', '70.6%', '262k'],
+  ['qwen/qwen3-coder:free',                     'Qwen3 Coder 480B',   'S+', '70.6%', '1M'],
   // Removed (2026-05-31): minimax/minimax-m2.5:free (free tier removed from OpenRouter)
   ['z-ai/glm-4.5-air:free',                     'GLM 4.5 Air',        'S+', '72.0%', '131k'],
   // Removed (2026-05-31): deepseek/deepseek-v4-flash:free (free tier removed from OpenRouter)
-  ['poolside/laguna-m.1:free',                  'Poolside Laguna M.1', 'S+', '-',     '131k'],
+  ['poolside/laguna-m.1:free',                  'Poolside Laguna M.1', 'S+', '-',     '262k'],
   ['poolside/laguna-xs.2:free',                 'Poolside Laguna XS.2','S+', '-',     '131k'],
   ['moonshotai/kimi-k2.6:free',                 'Kimi K2.6',          'S+', '76.8%', '262k'],
   // ── S tier — live :free chat/coding models ──
@@ -246,7 +246,9 @@ export const googleai = [
   ['gemini-3.1-flash-lite',                     'Gemini 3.1 Flash Lite',        'A+', '55.0%', '1M'],
   // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-pro',                            'Gemini 2.5 Pro',               'S+', '63.2%', '1M'],
+  // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-flash',                          'Gemini 2.5 Flash',             'A+', '50.0%', '1M'],
+  // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-flash-lite',                     'Gemini 2.5 Flash Lite',        'A',  '42.0%', '1M'],
   // Removed (2026-05-26): gemini-3.1-flash-lite-preview (endpoint shutdown May 25, 2026 → replaced by gemini-3.1-flash-lite)
 ]
@@ -293,20 +295,24 @@ export const cloudflare = [
   ['@cf/moonshotai/kimi-k2.6',                'Kimi K2.6',         'S+', '76.8%', '262k'],
   // ── S tier ──
   ['@cf/zai-org/glm-4.7-flash',               'GLM-4.7-Flash',     'S',  '59.2%', '131k'],
-  ['@cf/openai/gpt-oss-120b',                 'GPT OSS 120B',      'S',  '60.0%', '128k'],
   // ── A+ tier ──
   // Removed (2026-05-31): @cf/qwen/qwq-32b (upstream deprecation — Groq removed qwen-qwq-32b July 2025)
-  ['@cf/nvidia/nemotron-3-120b-a12b',         'Nemotron 3 Super',  'A+', '56.0%', '128k'],
+  ['@cf/nvidia/nemotron-3-120b-a12b',         'Nemotron 3 Super',  'A+', '56.0%', '32k'],
   // ── A tier ──
   ['@cf/meta/llama-4-scout-17b-16e-instruct', 'Llama 4 Scout',     'A',  '44.0%', '131k'],
+  ['@cf/meta/llama-3.2-90b-instruct',          'Llama 3.2 90B',     'A+', '45.0%', '128k'],
+  ['@cf/mistralai/mistral-7b-instruct-v0.2',   'Mistral 7B v0.2',   'A',  '38.0%', '128k'],
+  ['@cf/google/gemma-2-9b-it',                 'Gemma 2 9B',        'A',  '35.0%', '128k'],
+  ['@cf/anthropic/claude-3-5-sonnet',         'Claude 3.5 Sonnet',  'S+', '62.4%', '200k'],
+  ['@cf/openai/gpt-4o-mini',                   'GPT-4o Mini',       'A+', '50.0%', '128k'],
   ['@cf/qwen/qwen3-30b-a3b-fp8',              'Qwen3 30B MoE',     'A',  '45.0%', '128k'],
-  ['@cf/qwen/qwen2.5-coder-32b-instruct',     'Qwen2.5 Coder 32B', 'A',  '46.0%', '32k'],
+  ['@cf/qwen/qwen2.5-coder-32b-instruct',     'Qwen2.5 Coder 32B', 'A',  '46.0%', '128k'],
   ['@cf/openai/gpt-oss-20b',                  'GPT OSS 20B',       'A',  '42.0%', '128k'],
   // ── A- tier ──
   ['@cf/meta/llama-3.3-70b-instruct-fp8-fast','Llama 3.3 70B',     'A-', '39.5%', '128k'],
-  ['@cf/google/gemma-4-26b-a4b-it',           'Gemma 4 26B MoE',   'A-', '38.0%', '256k'],
+  ['@cf/google/gemma-4-26b-a4b-it',           'Gemma 4 26B MoE',   'A-', '38.0%', '128k'],
   ['@cf/mistralai/mistral-small-3.1-24b-instruct', 'Mistral Small 3.1', 'B+', '30.0%', '128k'],
-  ['@cf/deepseek-ai/deepseek-r1-distill-qwen-32b', 'DeepSeek R1 Distill 32B', 'A-', '45.0%', '80k'],
+
   // ── B tier ──
   ['@cf/ibm/granite-4.0-h-micro',             'Granite 4.0 Micro', 'B+', '30.0%', '128k'],
   // Removed (2026-05-26): @cf/google/gemma-4-31b-it (never existed on Cloudflare)
@@ -346,7 +352,9 @@ export const gemini = [
   ['gemini-3.1-flash-lite',             'Gemini 3.1 Flash Lite',        'A+', '55.0%', '1M'],
   // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-pro',                    'Gemini 2.5 Pro',               'S+', '63.2%', '1M'],
+  // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-flash',                  'Gemini 2.5 Flash',             'A+', '50.0%', '1M'],
+  // ⚠️ DEPRECATED — shutdown Oct 16, 2026
   ['gemini-2.5-flash-lite',             'Gemini 2.5 Flash Lite',        'A',  '42.0%', '1M'],
   // Removed (2026-05-26): gemini-3.1-flash-lite-preview (endpoint shutdown May 25, 2026)
 ]

package/src/core/utils.js

@@ -115,6 +115,82 @@ export const getAvg = (r) => {
 //
 // 📖 The "wasUpBefore" check is key — it distinguishes between a model that's
 //    temporarily flaky vs one that was never reachable in the first place.
+const NEW_MODELS = new Set([
+  'nvidia/nemotron-3-ultra-550b-a55b',
+  '@cf/meta/llama-3.2-90b-instruct',
+  '@cf/mistralai/mistral-7b-instruct-v0.2',
+  '@cf/google/gemma-2-9b-it',
+  '@cf/anthropic/claude-3-5-sonnet',
+  '@cf/openai/gpt-4o-mini',
+  '@cf/qwen/qwen3-30b-a3b-fp8',
+  '@cf/qwen/qwen2.5-coder-32b-instruct',
+  '@cf/meta/llama-3.3-70b-instruct-fp8-fast',
+  '@cf/google/gemma-4-26b-a4b-it',
+  '@cf/mistralai/mistral-small-3.1-24b-instruct',
+  '@cf/ibm/granite-4.0-h-micro',
+  'minimaxai/minimax-m2.7',
+  'z-ai/glm-5.1',
+  'moonshotai/kimi-k2.6',
+  'stepfun-ai/step-3.5-flash',
+  'stepfun-ai/step-3.7-flash',
+  'qwen/qwen3-coder-480b-a35b-instruct',
+  'qwen/qwen3.5-397b-a17b',
+  'meta/llama-4-maverick-17b-128e-instruct',
+  'mistralai/mistral-medium-3.5-128b',
+  'mistralai/mistral-small-4-119b-2603',
+  'qwen/qwen3.5-122b-a10b',
+  'mistralai/mistral-large-3-675b-instruct-2512',
+  'nvidia/nemotron-3-super-120b-a12b',
+  'nvidia/nemotron-3-nano-omni-30b-a3b-reasoning',
+  'google/gemma-4-31b-it',
+  'bytedance/seed-oss-36b-instruct',
+  'stockmark/stockmark-2-100b-instruct',
+  'mistralai/ministral-14b-instruct-2512',
+  'meta/llama-3.2-11b-vision-instruct',
+  'microsoft/phi-4-mini-instruct',
+  'gemma-3-12b-it',
+  'nvidia/nemotron-nano-9b-v2',
+  'openrouter/owl-alpha',
+  'nousresearch/hermes-3-llama-3.1-405b:free',
+  'nvidia/nemotron-nano-30b-a3b:free',
+  'cognitivecomputations/dolphin-mistral-24b-venice-edition:free',
+  'meta-llama/llama-3.3-70b-instruct:free',
+  'meta-llama/llama-3.2-3b-instruct:free',
+  'liquid/lfm-2.5-1.2b-instruct:free',
+  'liquid/lfm-2.5-1.2b-thinking:free',
+  'qwen3.7-max',
+  'qwen3-max',
+  'qwen3.6-plus',
+  'qwen3-235b-a22b',
+  'qwen3.5-plus',
+  'qwen3-coder-plus',
+  'qwen3-coder-next',
+  'qwen3.6-flash',
+  'qwen3.5-flash',
+  'qwen3-coder-flash',
+  'qwen3-32b',
+  'qwen3-coder-30b-a3b-instruct',
+  'holo2-30b-a3b',
+  'llama-3.3-70b-instruct',
+  'mistral-small-3.2-24b-instruct-2506',
+  'gemma-3-27b-it',
+  'qwen3.5-397b-a17b',
+  'qwen3-coder-30b-a3b-instruct',
+  'gpt-oss-120b',
+  'gpt-oss-20b',
+  'Meta-Llama-3_3-70B-Instruct',
+  'Qwen3-32B',
+  'Mistral-Small-3.2-24B-Instruct-2506',
+  'Mistral-7B-Instruct-v0.3',
+  'Mistral-Nemo-Instruct-2407',
+  'Qwen3.5-9B',
+  'big-pickle',
+  'deepseek-v4-flash-free',
+  'mimo-v2.5-free',
+  'nemotron-3-super-free',
+  'minimax-m3-free'
+]);
+
 export const getVerdict = (r) => {
   const avg = getAvg(r)
   const wasUpBefore = r.pings.length > 0 && r.pings.some(p => p.code === '200')
@@ -128,6 +204,34 @@ export const getVerdict = (r) => {
   const measurablePings = (r.pings || []).filter(p => measurablePingCodes.has(p.code))
   const p95 = getP95(r)
 
+  // 📖 Incorporate benchmark data (AI Latency and TPS) if available
+  if (r.benchmark && r.benchmark.ok) {
+    // AI Latency from benchmark (totalMs)
+    const aiLatency = r.benchmark.totalMs
+    // TPS from benchmark (tokens per second)
+    const tps = r.benchmark.tokensPerSecond
+    
+    // Adjust verdict based on benchmark data
+    if (aiLatency < 400) {
+      // 📖 Only flag as "Spiky" when we have enough data (≥3 pings) to judge stability
+      if (measurablePings.length >= 3 && p95 > 3000) return 'Spiky'
+      return 'Perfect'
+    }
+    if (aiLatency < 1000) {
+      if (measurablePings.length >= 3 && p95 > 5000) return 'Spiky'
+      return 'Normal'
+    }
+    if (aiLatency < 3000) return 'Slow'
+    if (aiLatency < 5000) return 'Very Slow'
+    if (aiLatency < 10000) return 'Unstable'
+    
+    // 📖 High TPS can improve verdict for models with higher latency
+    if (tps > 20 && aiLatency < 15000) return 'Slow'
+    if (tps > 40 && aiLatency < 20000) return 'Normal'
+    if (tps > 60 && aiLatency < 25000) return 'Perfect'
+  }
+  
+  // 📖 Fall back to ping-based verdict if no benchmark data
   if (avg < 400) {
     // 📖 Only flag as "Spiky" when we have enough data (≥3 pings) to judge stability
     if (measurablePings.length >= 3 && p95 > 3000) return 'Spiky'
@@ -236,7 +340,10 @@ export const getStabilityScore = (r) => {
 //
 // 📖 sortDirection 'asc' = ascending (smallest first), 'desc' = descending (largest first)
 export const sortResults = (results, sortColumn, sortDirection, { benchmarkResults = {} } = {}) => {
-  return [...results].sort((a, b) => {
+  return [...results].map(r => ({
+    ...r,
+    benchmark: benchmarkResults?.[`${r.providerKey}/${r.modelId}`]
+  })).sort((a, b) => {
     let cmp = 0
 
     switch (sortColumn) {

package/src/tui/render-table.js

@@ -794,66 +794,77 @@ export function renderTable({
     })() : statusText
     const status = statusColor(padEndDisplay(statusDisplayText, wStatus))
 
-    // 📖 Verdict column - use getVerdict() for stability-aware verdicts, then render with emoji
-    const verdict = getVerdict(r)
-    let verdictText, verdictIcon, verdictColor
-    // 📖 Verdict colors follow the same green→red gradient as TIER_COLOR / SWE%
-    switch (verdict) {
-      case 'Perfect':
-        verdictIcon = '🟩'
-        verdictText = `${verdictIcon} Perfect`
-        verdictColor = themeColors.successBold
-        break
-      case 'Normal':
-        verdictIcon = '🟢'
-        verdictText = `${verdictIcon} Normal`
-        verdictColor = themeColors.metricGood
-        break
-      case 'Spiky':
-        verdictIcon = '🟡'
-        verdictText = `${verdictIcon} Spiky`
-        verdictColor = (text) => chalk.bold.rgb(...getTierRgb('A+'))(text)
-        break
-      case 'Slow':
-        verdictIcon = '🟠'
-        verdictText = `${verdictIcon} Slow`
-        verdictColor = (text) => chalk.bold.rgb(...getTierRgb('A-'))(text)
-        break
-      case 'Very Slow':
-        verdictIcon = '🔴'
-        verdictText = `${verdictIcon} Very Slow`
-        verdictColor = (text) => chalk.bold.rgb(...getTierRgb('B+'))(text)
-        break
-      case 'Overloaded':
-        verdictIcon = '🔥'
-        verdictText = `${verdictIcon} Overloaded`
-        verdictColor = (text) => chalk.bold.rgb(...getTierRgb('B'))(text)
-        break
-      case 'Unstable':
-        // 📖 Avoid ⚠️ here: its variation selector has inconsistent terminal width and shifts the tiny ❔ column.
-        verdictIcon = '🟥'
-        verdictText = `${verdictIcon} Unstable`
-        verdictColor = themeColors.errorBold
-        break
-      case 'Not Active':
-        verdictIcon = '⚫'
-        verdictText = `${verdictIcon} Not Active`
-        verdictColor = themeColors.dim
-        break
-      case 'Pending':
-        verdictIcon = '⏳'
-        verdictText = `${verdictIcon} Pending`
-        verdictColor = themeColors.dim
-        break
-      default:
-        verdictIcon = '💀'
-        verdictText = `${verdictIcon} Unusable`
-        verdictColor = (text) => chalk.bold.rgb(...getTierRgb('C'))(text)
-        break
-    }
-    // 📖 Use padEndDisplay to account for emoji display width (2 cols each) so all rows align
-    const speedCell = verdictColor(padEndDisplay(verdictText, W_VERDICT))
-    const moodCell = padEndDisplay(verdictIcon, W_MOOD)
+  // 📖 Verdict column - use getVerdict() for stability-aware verdicts, then render with emoji
+  const verdict = getVerdict(r)
+  let verdictText, verdictIcon, verdictColor
+  // 📖 Verdict colors follow the same green→red gradient as TIER_COLOR / SWE%
+  switch (verdict) {
+    case 'Perfect':
+      verdictIcon = '🟩✨'
+      verdictText = `${verdictIcon} Perfect`
+      verdictColor = themeColors.successBold
+      break
+    case 'Normal':
+      verdictIcon = '🟢✨'
+      verdictText = `${verdictIcon} Normal`
+      verdictColor = themeColors.metricGood
+      break
+    case 'Spiky':
+      verdictIcon = '🟡✨'
+      verdictText = `${verdictIcon} Spiky`
+      verdictColor = (text) => chalk.bold.rgb(...getTierRgb('A+'))(text)
+      break
+    case 'Slow':
+      verdictIcon = '🟠✨'
+      verdictText = `${verdictIcon} Slow`
+      verdictColor = (text) => chalk.bold.rgb(...getTierRgb('A-'))(text)
+      break
+    case 'Very Slow':
+      verdictIcon = '🔴✨'
+      verdictText = `${verdictIcon} Very Slow`
+      verdictColor = (text) => chalk.bold.rgb(...getTierRgb('B+'))(text)
+      break
+    case 'Overloaded':
+      verdictIcon = '🔥✨'
+      verdictText = `${verdictIcon} Overloaded`
+      verdictColor = (text) => chalk.bold.rgb(...getTierRgb('B'))(text)
+      break
+    case 'Unstable':
+      // 📖 Avoid ⚠️ here: its variation selector has inconsistent terminal width and shifts the tiny ❔ column.
+      verdictIcon = '🟥✨'
+      verdictText = `${verdictIcon} Unstable`
+      verdictColor = themeColors.errorBold
+      break
+    case 'Not Active':
+      verdictIcon = '⚫✨'
+      verdictText = `${verdictIcon} Not Active`
+      verdictColor = themeColors.dim
+      break
+    case 'Pending':
+      verdictIcon = '⏳✨'
+      verdictText = `${verdictIcon} Pending`
+      verdictColor = themeColors.dim
+      break
+    default:
+      verdictIcon = '💀✨'
+      verdictText = `${verdictIcon} Unusable`
+      verdictColor = (text) => chalk.bold.rgb(...getTierRgb('C'))(text)
+      break
+  }
+  // 📖 Use padEndDisplay to account for emoji display width (2 cols each) so all rows align
+  const speedCell = verdictColor(padEndDisplay(verdictText, W_VERDICT))
+  const moodCell = padEndDisplay(verdictIcon, W_MOOD)
+  
+  // 📖 Add NEW badge for recently added models
+  const isNewModel = NEW_MODELS.has(r.modelId) || NEW_MODELS.has(`${r.providerKey}/${r.modelId}`)
+  const newBadge = isNewModel ? '🆕' : ''
+  const verdictTextWithBadge = newBadge ? `${newBadge} ${verdictText}` : verdictText
+  const speedCellWithBadge = verdictColor(padEndDisplay(verdictTextWithBadge, W_VERDICT))
+  const moodCellWithBadge = padEndDisplay(verdictIcon, W_MOOD)
+  
+  // 📖 Use the version with badge for display
+  const speedCell = speedCellWithBadge
+  const moodCell = moodCellWithBadge
 
     // 📖 Stability column - composite score (0–100) from p95 + jitter + spikes + uptime
     // 📖 Left-aligned to sit flush under the column header