feat(benchmark): update Sonnet 5.5 / Opus 5.5 pricing and repoint aliases

cmd/cc-session/benchmark.go

@@ -30,7 +30,7 @@ func runBenchmark(args []string, out io.Writer, errOut io.Writer, store parser.S
 	days := fs.Int("days", 30, "how far back to scan")
 	minKB := fs.Int("min-kb", 100, "minimum JSONL file size in KB")
 	maxN := fs.Int("n", 10, "max sessions to include")
-	model := fs.String("model", "opus", "model: opus, opus-4-6, opus-4-7, opus-4-8, opus-5-5, sonnet, or fable (fable-5-1)")
+	model := fs.String("model", "opus", "model: opus (opus-5-5), opus-4-6, opus-4-7, opus-4-8, sonnet (sonnet-5-5), sonnet-4-6, or fable (fable-5-1)")
 	overhead := fs.Int("overhead", 0, "session overhead tokens (system+tools+CLAUDE.md); measure with a 1-turn session")
 	isNoAPI := fs.Bool("no-api", false, "skip API calls; estimate filtered-text tokens with chars/2 (offline fallback)")
 	if err := fs.Parse(reorderArgs(args)); err != nil {

cmd/cc-session/benchmark_test.go

@@ -81,11 +81,12 @@ func TestResolveBenchmarkModel_GivenAcceptedModelNames_ThenReturnsPricingAndToke
 		wantPricing         bm.Pricing
 		wantTokenCountModel string
 	}{
-		{name: "opus", wantPricing: bm.PricingOpus, wantTokenCountModel: bm.TokenCountModelOpus48},
+		{name: "opus", wantPricing: bm.PricingOpus55, wantTokenCountModel: bm.TokenCountModelOpus55},
 		{name: "opus-4-6", wantPricing: bm.PricingOpus, wantTokenCountModel: bm.TokenCountModelOpus46},
 		{name: "opus-4-7", wantPricing: bm.PricingOpus, wantTokenCountModel: bm.TokenCountModelOpus47},
 		{name: "opus-4-8", wantPricing: bm.PricingOpus, wantTokenCountModel: bm.TokenCountModelOpus48},
-		{name: "sonnet", wantPricing: bm.PricingSonnet, wantTokenCountModel: bm.TokenCountModelSonnet},
+		{name: "sonnet", wantPricing: bm.PricingSonnet55, wantTokenCountModel: bm.TokenCountModelSonnet55},
+		{name: "sonnet-4-6", wantPricing: bm.PricingSonnet46, wantTokenCountModel: bm.TokenCountModelSonnet46},
 	}
 
 	for _, tt := range tests {
@@ -111,7 +112,7 @@ func TestResolveBenchmarkModel_GivenUnknownModel_ThenReturnsAcceptedNames(t *tes
 	}
 
 	got := err.Error()
-	for _, want := range []string{"opus", "opus-4-6", "opus-4-7", "opus-4-8", "sonnet"} {
+	for _, want := range []string{"opus", "opus-4-6", "opus-4-7", "opus-4-8", "sonnet", "sonnet-4-6"} {
 		if !strings.Contains(got, want) {
 			t.Fatalf("error message missing accepted model %q: %s", want, got)
 		}
@@ -171,7 +172,7 @@ func TestRunBenchmark_WhenSessionHasAPIUsage_ThenUsesTokenCountingAPIForNewConte
 	if countedText == "" {
 		t.Fatal("countTokensFn was not called")
 	}
-	wantCountModel := bm.TokenCountModelOpus48
+	wantCountModel := bm.TokenCountModelOpus55
 	if countModel != wantCountModel {
 		t.Fatalf("token counter model = %q, want %q", countModel, wantCountModel)
 	}
@@ -328,7 +329,7 @@ func TestRunBenchmark_GivenSonnetModel_ThenUsesSonnetTokenCounterModel(t *testin
 		t.Fatalf("runBenchmark returned error: %v", err)
 	}
 
-	wantCountModel := bm.TokenCountModelSonnet
+	wantCountModel := bm.TokenCountModelSonnet55
 	if countModel != wantCountModel {
 		t.Fatalf("token counter model = %q, want %q", countModel, wantCountModel)
 	}

docs/benchmark.md

@@ -36,7 +36,7 @@ All other parameters are derived automatically from your real session data.
 | `--days` | 30 | How far back to scan for sessions |
 | `--min-kb` | 100 | Minimum JSONL file size in KB |
 | `--n` | 10 | Max successful session results to report |
-| `--model` | opus | Pricing and token-counting model: `opus`, `opus-4-6`, `opus-4-7`, `opus-4-8`, `opus-5-5`, `sonnet`, or `fable` (`fable-5-1`) |
+| `--model` | opus | Pricing and token-counting model: `opus` (`opus-5-5`), `opus-4-6`, `opus-4-7`, `opus-4-8`, `sonnet` (`sonnet-5-5`), `sonnet-4-6`, or `fable` (`fable-5-1`) |
 
 ### Example output
 
@@ -162,17 +162,19 @@ the historical one-shot `NewCtx × CacheWrite` behavior.
 
 `X` comes from transcript API usage and `C` comes from the Anthropic token
 counting API. The `--model` flag controls both pricing and the tokenizer used by
-the token counting API. `opus` is an alias for `opus-4-8`; explicit Opus 4.x
-versions map to `claude-opus-4-6`, `claude-opus-4-7`, or `claude-opus-4-8`;
-`opus-5-5` maps to `claude-opus-5-5`; `sonnet` maps to `claude-sonnet-4-6`;
+the token counting API. `opus` and `opus-5-5` both map to `claude-opus-5-5`;
+explicit Opus 4.x versions map to `claude-opus-4-6`, `claude-opus-4-7`, or
+`claude-opus-4-8`. `sonnet` and `sonnet-5-5` both map to `claude-sonnet-5-5`
+($2.00/M input, $0.20/M cache read, $2.50/M cache write, $10/M output);
+`sonnet-4-6` maps to `claude-sonnet-4-6`.
 `fable` and `fable-5-1` both map to `claude-fable-5-1`.
 Opus 4.6, 4.7, and 4.8 use the same Opus pricing rates; Opus 5.5 has its own,
 lower rates (see the pricing table above).
 Fallback constants are used only for behavior that cannot be read directly from
 transcript usage, such as sparse tool I/O data.
 
 Cache read as a percentage of base input differs by model: Opus 4.x and Sonnet
-use 10%, Opus 5.5 uses 5%, and Claude Fable 5.1 uses 2.5%. Cost-savings numbers
+(4.6 and 5.5) use 10%, Opus 5.5 uses 5%, and Claude Fable 5.1 uses 2.5%. Cost-savings numbers
 are only directly comparable across rows using the same ratio.
 
 ### Simplifications

internal/benchmark/model.go

@@ -9,8 +9,13 @@ type Pricing struct {
 	BaseInput  float64 // $/M tokens (uncached, after last breakpoint)
 }
 
+// PricingOpus is the shared Opus 4.x pricing (4.6, 4.7, 4.8).
 var PricingOpus = Pricing{CachedRead: 0.50, CacheWrite: 6.25, BaseInput: 5.00}
-var PricingSonnet = Pricing{CachedRead: 0.30, CacheWrite: 3.75, BaseInput: 3.00}
+var PricingSonnet46 = Pricing{CachedRead: 0.30, CacheWrite: 3.75, BaseInput: 3.00}
+
+// PricingSonnet55 is Claude Sonnet 5.5's pricing ($2 input / $10 output per
+// MTok). Cache read stays at 10% of base input, like Sonnet 4.6.
+var PricingSonnet55 = Pricing{CachedRead: 0.20, CacheWrite: 2.50, BaseInput: 2.00}
 
 // PricingOpus55 is Claude Opus 5.5's pricing. Its cache read is 5% of base
 // input rather than the 10% Opus 4.x/Sonnet use (source: Anthropic pricing
@@ -26,12 +31,13 @@ var PricingOpus55 = Pricing{CachedRead: 0.20, CacheWrite: 5.00, BaseInput: 4.00}
 var PricingFable = Pricing{CachedRead: 0.25, CacheWrite: 12.50, BaseInput: 10.00}
 
 const (
-	TokenCountModelOpus46 = "claude-opus-4-6"
-	TokenCountModelOpus47 = "claude-opus-4-7"
-	TokenCountModelOpus48 = "claude-opus-4-8"
-	TokenCountModelOpus55 = "claude-opus-5-5"
-	TokenCountModelSonnet = "claude-sonnet-4-6"
-	TokenCountModelFable  = "claude-fable-5-1"
+	TokenCountModelOpus46   = "claude-opus-4-6"
+	TokenCountModelOpus47   = "claude-opus-4-7"
+	TokenCountModelOpus48   = "claude-opus-4-8"
+	TokenCountModelOpus55   = "claude-opus-5-5"
+	TokenCountModelSonnet46 = "claude-sonnet-4-6"
+	TokenCountModelSonnet55 = "claude-sonnet-5-5"
+	TokenCountModelFable    = "claude-fable-5-1"
 )
 
 // ModelConfig bundles the pricing and token-counting model for a given model alias.
@@ -43,20 +49,22 @@ type ModelConfig struct {
 // ResolveModel maps a user-facing model alias to its ModelConfig.
 func ResolveModel(model string) (ModelConfig, error) {
 	switch model {
-	case "sonnet":
-		return ModelConfig{Pricing: PricingSonnet, TokenCountModel: TokenCountModelSonnet}, nil
-	case "opus", "opus-4-8":
+	case "sonnet", "sonnet-5-5":
+		return ModelConfig{Pricing: PricingSonnet55, TokenCountModel: TokenCountModelSonnet55}, nil
+	case "sonnet-4-6":
+		return ModelConfig{Pricing: PricingSonnet46, TokenCountModel: TokenCountModelSonnet46}, nil
+	case "opus", "opus-5-5":
+		return ModelConfig{Pricing: PricingOpus55, TokenCountModel: TokenCountModelOpus55}, nil
+	case "opus-4-8":
 		return ModelConfig{Pricing: PricingOpus, TokenCountModel: TokenCountModelOpus48}, nil
 	case "opus-4-7":
 		return ModelConfig{Pricing: PricingOpus, TokenCountModel: TokenCountModelOpus47}, nil
 	case "opus-4-6":
 		return ModelConfig{Pricing: PricingOpus, TokenCountModel: TokenCountModelOpus46}, nil
-	case "opus-5-5":
-		return ModelConfig{Pricing: PricingOpus55, TokenCountModel: TokenCountModelOpus55}, nil
 	case "fable", "fable-5-1":
 		return ModelConfig{Pricing: PricingFable, TokenCountModel: TokenCountModelFable}, nil
 	default:
-		return ModelConfig{}, fmt.Errorf("unknown model %q: must be opus, opus-4-6, opus-4-7, opus-4-8, opus-5-5, sonnet, fable, or fable-5-1", model)
+		return ModelConfig{}, fmt.Errorf("unknown model %q: must be opus, opus-4-6, opus-4-7, opus-4-8, opus-5-5, sonnet, sonnet-4-6, sonnet-5-5, fable, or fable-5-1", model)
 	}
 }
 

internal/benchmark/model_test.go

@@ -15,14 +15,26 @@ func Test_ResolveModel_GivenKnownAlias_ThenReturnsExpectedConfig(t *testing.T) {
 		{
 			name:           "sonnet",
 			alias:          "sonnet",
-			wantPricing:    PricingSonnet,
-			wantTokenCount: TokenCountModelSonnet,
+			wantPricing:    PricingSonnet55,
+			wantTokenCount: TokenCountModelSonnet55,
+		},
+		{
+			name:           "sonnet-4-6",
+			alias:          "sonnet-4-6",
+			wantPricing:    PricingSonnet46,
+			wantTokenCount: TokenCountModelSonnet46,
+		},
+		{
+			name:           "sonnet-5-5",
+			alias:          "sonnet-5-5",
+			wantPricing:    PricingSonnet55,
+			wantTokenCount: TokenCountModelSonnet55,
 		},
 		{
 			name:           "opus",
 			alias:          "opus",
-			wantPricing:    PricingOpus,
-			wantTokenCount: TokenCountModelOpus48,
+			wantPricing:    PricingOpus55,
+			wantTokenCount: TokenCountModelOpus55,
 		},
 		{
 			name:           "opus-4-8",

internal/tokens/tokens.go

@@ -29,7 +29,7 @@ const (
 	baseRetryDelay = 500 * time.Millisecond
 )
 
-const DefaultCountTokensModel = "claude-sonnet-4-6"
+const DefaultCountTokensModel = "claude-sonnet-5-5"
 
 // CountTokensAPI calls the Anthropic count_tokens endpoint.
 // Resolves the API key from: env ANTHROPIC_API_KEY → config file path in